MCP projects built at Amrita University Coimbatore
These are 183 real MCP apps and servers built and deployed by student teams at the Amrita University Coimbatore NitroStack × MCP To The Moon hackathon. Each project is a working agentic AI build on the Model Context Protocol, shipped on NitroStack. Open any project to watch the demo, read the write-up, and explore the source.
Rightly is an intelligent Model Context Protocol (MCP) server that acts as a personal consumer protection advocate. Pre-purchase, it performs deep analysis on product listings by extracting true specifications, filtering out fake reviews, assessing repairability scores, and identifying deceptive marketing. Post-purchase, it generates step-by-step resolution plans for defective goods. When disputes arise, Rightly can instantly locate hard-to-find official support channels, find local repair centers, and automatically draft and email formal legal demand notices to vendors.
Who it is for: It is built for everyday consumers who want to make highly informed purchasing decisions and avoid buyer's remorse, as well as those feeling overwhelmed by confusing return policies or unresponsive customer support when dealing with defective products.
What makes it special: Rightly bridges the gap between AI chat and interactive UI. Instead of just returning walls of text, it uses the NitroStack Framework to render beautiful, interactive React widgets directly inside the AI client. Users can review rich product analysis cards, click through interactive resolution checklists, review drafted legal documents on a formal UI, and click a single "Send via Email" button to dispatch their legal notice directly to the seller—all without ever leaving the chat interface. It also features intelligent offline fallbacks to ensure consumers always have access to support links, even if APIs fail.
Vayu is a Model Context Protocol (MCP) server that transforms an AI into a proactive sun-safety advisor. By equipping an LLM with specialized environmental and dermatological tools, Vayu can:
- Fetch Real-Time UV Data : Look up the current UV index for any global coordinates.
- Calculate Safe Exposure Windows : Determine exactly how many minutes a user can safely stay in the sun before risking skin damage.
- Track Risk Escalation : Monitor changes in the UV index throughout the day and trigger alerts if the environmental risk jumps to a dangerous tier.
- Manage User Profiles : Store and recall specific user skin types (based on the Fitzpatrick scale 1-6) to ensure all advice is deeply personalized.
Who is it for?
- Outdoor Enthusiasts : Hikers, runners, beachgoers, or travelers who spend prolonged periods outside and need precise guidelines on when to seek shade or reapply sunscreen.
- Sun-Sensitive Individuals : People with skin conditions, a history of skin cancer, or those using photosensitizing medications who need strict, reliable environmental monitoring.
- Parents & Caregivers : People wanting to ensure their children are safe during outdoor play or sports events.
What makes it special?
- Hyper-Personalization : Unlike generic weather apps that just output a number, Vayu cross-references the raw UV index with the user's specific skin type to calculate an exact, actionable "safe exposure window" in minutes.
- Proactive State Tracking : Vayu maintains session state across interactions. It doesn't just answer one-off questions; it can track previous risk tiers and alert the AI if the environment has suddenly escalated (e.g., jumping from "Moderate" to "Extreme").
- Seamless AI Integration : As an MCP server, Vayu allows an LLM to seamlessly weave sun safety into broader conversations. If you tell your AI, *"I'm going for a 3-hour hike in Denver at noon,"* the AI can automatically fetch the UV data, referenc
VeriCite is an AI-powered citation integrity auditor that automatically verifies whether a document's claims are accurately supported by their cited sources. Built for researchers, students, educators, publishers, and organizations using AI-assisted writing, it detects fabricated or weak citations, measures citation coverage, analyzes evidence quality, and generates an intuitive trust score with detailed explanations. By combining multi-source scholarly verification, contradiction detection, and automated evidence retrieval, VeriCite helps users identify misinformation before publication, significantly reducing manual verification time while improving the reliability and credibility of technical and academic documents.
PolyRisk transforms raw genetic data into clinically grounded insight, cross-referencing an individual's genome against real, published GWAS research across 34 diseases and 59 validated variants.
Every result is traced to a peer-reviewed study via live citation, ensuring complete scientific transparency with zero fabricated data.
Built as an agentic MCP pipeline, the system autonomously evaluates the strength of evidence and statistical significance before surfacing any findings.
PolyRisk is a research-grade prototype, not a diagnostic tool, designed to demonstrate how accessible, evidence-based genetic risk interpretation can be
When a new business wants to accept card payments or borrow money, it fills in forms, uploads bank statements, and then waits **three to five business days** while someone at the bank reads those statements by hand.
That wait is expensive for everyone. The business cannot trade. The bank pays skilled analysts to do repetitive work. And because the review is manual, two analysts looking at the same statements can reach different conclusions — with no written record of why.
The evidence needed to make the decision is already sitting in the business's bank account. The delay is not analysis; it is queueing.
## 2. What InstantPulse does
The business securely connects its bank account. InstantPulse then:
1. Pulls the transaction history
2. Analyses cash flow, income stability, expenses, debt patterns and unusual transactions
3. Produces a **transparent risk score** out of 100, with a written reason for every point awarded or withheld
4. Recommends a **credit limit**, and says which constraint capped it
5. Classifies the application:
- 🟢 **Green** — ready to proceed
- 🟡 **Yellow** — human review required
- 🔴 **Red** — high risk or missing information
6. For Green applications, automatically starts **Stripe** payment-account onboarding
7. Gives managers a live dashboard explaining the decision, the risk factors and the next action
Three to five days becomes **one call**. In offline mode a full decision takes about 30 milliseconds; against the live Plaid API it takes about 15 seconds.
Crucially, it does **not** replace credit officers. Yellow applications go into a real review queue where an officer can request documents or override the decision — with a written justification that is recorded permanently.
This project, MedGuard App, is a clinical medication safety evaluation server built on the NitroStack Model Context Protocol (MCP) framework. It acts as an intelligent assistant to help healthcare professionals and AI models cross-reference proposed medications against patients' medical histories.
Here is a summary of the project's architecture and capabilities:
🛠️ Core Technology Stack
Framework: NitroStack (@nitrostack/core and @nitrostack/cli) using TypeScript and Zod schema validation.
Database: PostgreSQL / Supabase integration (@supabase/supabase-js, pg client) with automatic fallback to local JSON fixtures (patients.json & medguard-dataset.json).
Development Tooling: Configured with tsx for script execution and TypeScript compiling.
🛡️ Key Features & Tools
1. Medication Safety Evaluation (medguard.tools.ts)
evaluate_treatment_safety: Cross-checks a proposed drug (e.g., dermatology drug) against a patient's profile.
Checks for drug-disease or drug-drug contraindications.
Identifies severity/risk level (High, Medium, Low).
Triggers CRITICAL warnings if the risk is high and the patient's kidney function (eGFR < 60) is compromised.
Returns alternative treatments if safety thresholds are breached.
get_patient_details: Fetches clinical metrics (conditions, eGFR, active medication list) for a patient.
add_patient: Inserts a new patient record into either the live Supabase database or the local fallback cache.
list_interaction_rules: Allows searching the dataset of clinical interaction rules.
2. Database & Schema Initialization (database.tools.ts)
setup_medguard_database: Automates table creation and seeds initial patient/interaction rules into Supabase/PostgreSQL using
supabase-schema.sql
.
Supabase Schema: Exposes the database structure as an MCP Resource.
3. Standardized Workflows & Prompts
process_patient_workflow: A prompt template guiding the agent to verify patients, safely add new ones, inspect active drugs, and perform the safety evaluation.
medguard_q
Strat-OS is an AI-powered enterprise strategy platform that helps organizations analyze their business, identify growth opportunities, assess risks, and generate actionable strategic recommendations. Using an MCP-based architecture, it combines structured business data with AI reasoning to automate market analysis, strategic planning, compliance insights, and decision support.
The platform enables business leaders, consultants, and startups to create comprehensive enterprise strategies in minutes instead of weeks. It integrates multiple AI agents and enterprise knowledge sources to deliver accurate, explainable, and scalable recommendations while maintaining a modular architecture for future integrations.
Key features:
->AI-generated enterprise strategy reports
->Business and market analysis
->Risk and compliance insights
->Multi-agent workflow using MCP
->Scalable enterprise-ready architecture
->Fast, explainable decision support for organizations
AI CEO
Overview
AI CEO is an intelligent executive decision platform that acts as an always-on strategic advisor for organizations. It helps business leaders understand what is happening, why it is happening, and what actions should be taken by analyzing information across multiple business functions.
Unlike traditional dashboards that only display reports and metrics, AI CEO connects information from different departments, identifies relationships, uncovers trends, detects risks, and delivers evidence-backed insights. It is designed to assist human leadership, not replace it.
The Problem:
Organizations generate vast amounts of data across Engineering, Product, Sales, Finance, Customer Support, HR, Operations, and Project Management. Since this information is spread across different systems, leaders often struggle to gain a unified view of the organization, identify risks early, and understand how one business function impacts another.
The Solution:
AI CEO creates a centralized intelligence layer that continuously analyzes organizational information and transforms it into actionable business intelligence. It connects business activities, detects patterns, investigates changes, and provides contextual recommendations that support strategic decision-making.
Key Features:
Organization-wide business intelligence
Cross-functional analysis
Trend and anomaly detection
Risk identification
Executive summaries
Natural language interaction
Predictive business insights
Evidence-backed recommendations
Business Impact
AI CEO enables faster decision-making, improved organizational visibility, proactive risk management, and stronger strategic planning. Rather than simply reporting what happened, it explains why it happened, who is affected, and what actions should be considered, empowering leadership teams with connected, explainable, and actionable business intelligence.
Sprints and hackathons run on meetings. We discuss dependencies, schedule demos, delegate tasks, and identify blocks. But once the call ends, the execution falls apart. Action items are forgotten, deadlines slip, and key risks discussed during the meeting are never tracked—leading to missed launch dates and administrative overhead.
SymBioForge is an AI-powered platform that automatically connects factories so that one factory's waste becomes another factory's raw material. Think of it as a matchmaker for industrial waste. Imagine a textile mill that produces cotton lint as waste. A nearby paper mill could use that cotton lint as raw material. SymBioForge finds these connections automatically, calculates how much CO2 is saved, how much money each factory earns, and even generates government compliance reports.
An interactive black hole visualizer where a NitroStack MCP server computes real spacetime curvature (Christoffel symbols via SymPy) and a WebGPU frontend ray-traces actual light-bending in real time. Students and enthusiasts learning general relativity, and anyone curious what MCP servers can do beyond typical API-wrapper tools. The light-bending is real physics (RK4-integrated null geodesics), not a faked shader effect — and the underlying tensor math is exposed as verified, inspectable MCP tools rather than buried inside the renderer.
ContextOS is an Adaptive AI Context Intelligence Platform that transforms conversations into structured documentation, intelligent planning, workflow automation, and organizational knowledge. Unlike traditional AI meeting assistants, ContextOS does not automatically classify meetings. Instead, users manually select a Context Pack (Business, Software Development, Research, Healthcare, Education, Finance, Hackathon, and more), allowing the same conversation to generate entirely different workflows tailored to each domain.
Projectpilot AI is an AI-powered Project Manager and HR Manager that transforms a Software Requirements Document (SRD) into a complete execution plan. It automatically recommends the best SDLC model, assigns roles based on employee skills, creates day-wise, week-wise, and month-wise task schedules, and tracks GitHub contributions to measure productivity. Built with MCP and multi-agent AI, it helps startups, software teams, and student developers plan smarter, execute faster, and deliver projects on time
ClinicaMind is an AI-powered clinical decision support platform that streamlines patient intake and assists healthcare professionals in making faster, more informed decisions. It automatically retrieves patient intake forms from Gmail, processes medical documents using OCR and AI, extracts structured clinical information, and presents it in an intuitive workspace for doctor review. By combining intelligent document processing, AI-assisted risk insights, and a centralized patient workflow, ClinicaMind reduces administrative burden, minimizes manual errors, and enables clinicians to focus on delivering better patient care. It is built on NitroStack's MCP architecture with external Gmail API integration and an end-to-end healthcare workflow.
An MCP server that gives an AI agent everything it needs to run marketing campaigns. It posts and tracks analytics across Facebook, Instagram, and LinkedIn, sends and reads emails through Gmail, and connects to Google Drive and Slack to search, list, and read files, including converting Docs and Sheets into usable text or CSV. One server, all the tools an AI needs to plan, publish and manage a campaign.
LexoNova helps workers, contract employees, and gig workers understand whether what happened is a legal violation—and if so, what to do about it.
It doesn't just answer questions. It runs a full autonomous legal triage pipeline: identifying the issue, checking applicable law, flagging filing deadlines, determining the right authority to approach, and generating a ready-to-use case document—all from a single plain-language description of what happened.
Built for the NitroStack MCP Agentic AI Hackathon at Amrita Coimbatore.
AI-Powered Rural Digital Assistance Platform
Our project is an AI-powered digital assistant that helps rural and economically disadvantaged communities access essential services through a single multilingual platform. It provides guidance on financial assistance, government schemes, weather and soil analysis, free educational resources, government job opportunities, scam detection, and communication services like email and WhatsApp. The platform uses a hybrid AI architecture with a lightweight offline local AI model and cloud AI to ensure reliable assistance even in areas with poor or no internet connectivity. By combining voice support, multilingual interaction, and real-time public resources, the system aims to bridge the digital divide and improve accessibility for underserved communities across India.
An agentic ai model which contains tools to create a customized daily routine and a semester planner with inputs of the course curriculum which gives a planner for the student to follow
Every day, individuals and teams sign contracts blindly because legal review is prohibitively expensive, while cloud AI models expose confidential data and trade secrets.
Phalanx bridges this gap. Built on the open Model Context Protocol, it equips any AI assistant with policy-driven redaction, a Clause Knowledge Graph, and multi-agent risk analysis across Corporate, Financial, Liability, and Privacy domains. It classifies diverse contracts, uncovers cross-clause risks, benchmarks terms, and generates a negotiation email with fair redlines—all while keeping secrets in an encrypted local vault.
TARGET AUDIENCE
Individuals and Freelancers: Review leases and agreements without hourly legal fees.
Startups and Founders: Negotiate contracts without slowing down deal velocity.
Enterprise Legal and HR: Automate contract triage and enforce playbook compliance.
AI Developers: Embed secure legal intelligence into custom copilots and workflows.
INTELLIGENCE PIPELINE
Classification: Automatically detects contract categories and applies domain rules.
Policy-Driven Redaction: Anonymizes sensitive personal and commercial terms locally.
Clause Knowledge Graph: Maps dependencies so hidden obligations are never missed.
Multi-Agent Risk Engine: Concurrently audits governance, financial exposure, liability caps, and privacy compliance.
Actionable Synthesis: Delivers an overall risk score, fair redlines, and a negotiation email.
SOCIAL IMPACT
Democratizing Legal Protection: Protects renters, workers, and small businesses against predatory terms.
Privacy as a Fundamental Right: Sets a new ethical standard for AI in regulated sectors.
FINANCIAL IMPACT
Significant Cost Reduction: Replaces expensive preliminary legal reviews with instant audits.
Liability Prevention: Surfaces uncapped indemnities and silent auto-renewal traps before execution.
Accelerated Deal Velocity: Eliminates legal review bottlenecks to speed up time-to-revenue.
CREED Lite AI is an autonomous NitroStack MCP server that transforms traditional loan underwriting into a transparent, explainable, and fraud-safe decision support system. Instead of issuing binary approvals or rejections, it evaluates applicants using a continuous underwriting model, identifies the factors affecting eligibility, and generates personalized, budget-aware recourse plans to improve approval chances.
The platform combines AI-driven underwriting, causal recourse generation, fraud detection, and ECOA Regulation B compliance into a unified workflow. It explains rejection reasons, recommends safe financial interventions, validates them against fraud vectors, and simulates their impact on future approval probability. Interactive NitroStack UI widgets provide visual insights into applicant evaluation, legal disclosures, intervention strategies, and simulated outcomes.
By replacing opaque credit decisions with actionable, compliant, and explainable guidance, CREED-Lite empowers both lenders and applicants to make better financial decisions.
. What does it do?
HELIX is an AI-powered cognitive engine that acts as a real-time nervous system for an entire company. It ingests messy, siloed data from platforms like Jira, Slack, and Confluence, and visualizes the company as a living 3D "Genome." It automatically calculates a "Drift Score" for every department to detect when teams are violating policies or falling out of alignment. Most importantly, when a department reaches a severe drift threshold, HELIX uses autonomous AI to generate and execute real-world interventions—such as firing off executive alerts or restructuring workflows—to physically fix the problem.
2. Who is it for?
HELIX is built for Enterprise Executives, Chief Operating Officers (COOs), and Risk Officers at large-scale organizations. These leaders are currently flying blind; they manage thousands of employees across dozens of departments but only find out about massive inefficiencies, security risks, or compliance violations after they happen. HELIX is for leaders who want to stop looking at static charts of past data and start actively steering a self-healing organization.
3. What makes it special?
Three major technical breakthroughs make HELIX completely unique:
GraphRAG Architecture: Instead of a basic vector database, it uses GraphRAG (Retrieval-Augmented Generation on Graphs) to map the complex, topological relationships between people, projects, and goals, allowing the AI to understand the root cause of problems.
Model Context Protocol (MCP): We didn't just build a dashboard; we built a NitroStack MCP backend. This turns our Python analytics into autonomous AI "Tools" and "Resources" that can actively trigger workflows, rather than just displaying numbers.
From Passive to Active: Traditional Business Intelligence (BI) tools are read-only. HELIX is "read-write." It doesn't just tell you that Engineering is drifting; the built-in Intervention Studio allows you to chat with the AI and execute a targeted correction with a single click.
It is based on Industry 4.0 , This is intrusion detection and intrusion prevention system which detects the network traffic as normal or attack and The AI Agent takes prevention step automatically based on the detection but for critical attacks it verifies with a human (Man in the loop) before taking an decision. This uses Nitrostack Chatbot to retrieve data and generate reports and to view attack or approval request records and uses command line interface to track live network traffic.
Auditor Zero is an autonomous cross-document reasoning agent that catches what human review misses: contradictions between an organization's own documents, and things quietly dropped between versions. Search tools can tell you what a document says right now, they can't tell you what it used to say, why it stopped saying it, or what it silently contradicts elsewhere. Auditor Zero closes that gap.
It runs four complementary detectors in parallel: deterministic numeric-conflict checks, LLM-driven semantic comparison, cross-version change detection, and category-disappearance detection, each finding tagged with exactly how it was surfaced. Documents are grouped into version lineages by ingestion order (not fragile version-string parsing), so consecutive versions get diffed for weakened or vanished obligations, while different documents get compared pairwise for contradictions.
What sets it apart is the Black Box: every agent decision is sealed into a cryptographically keyed hash chain using HMAC-SHA256, not a plain hash. That distinction matters, a plain hash chain only proves something was logged in order; anyone editing a record can recompute a fresh, internally consistent chain. Auditor Zero's chain is keyed with a server-held secret, so no forged chain can ever reproduce the stored hashes. verify_replay_chain recomputes every link from genesis and reports the exact point where tampering occurred, if any.
Every operation, ingestion, contradiction detection, severity scoring, replay, verification is defined once and exposed identically as both an MCP tool and a REST endpoint, so the same reasoning pipeline a human clicks through in the dashboard is exactly what an autonomous agent calls under the hood.
Built on NitroStack in 24 hours: TypeScript, Express, the MCP SDK, Claude for reasoning, SQLite, JWT auth, and a full React dashboard with live-streaming audit progress and a tamper-detection demo mode.
EchoMind is an AI-powered engineering investigation platform built using NitroStack and the Model Context Protocol (MCP) to help software engineers resolve production incidents faster.
During production outages, engineers often spend valuable time switching between logs, source code, engineering documentation, and previous incident reports. EchoMind streamlines this workflow by providing intelligent tools to create investigation cases, collect code evidence, search engineering documentation, retrieve similar historical incidents, and organize investigation evidence within a single platform.
EchoMind is designed for software engineers, DevOps teams, Site Reliability Engineers (SREs), and engineering support teams who need to perform rapid root cause analysis during critical production failures.
Built on NitroStack's MCP ecosystem, EchoMind combines AI-assisted investigation, structured evidence management, and interactive widgets to reduce context switching and accelerate incident resolution. Instead of relying on multiple disconnected tools, engineering teams can perform the complete investigation workflow from one intelligent interface, improving collaboration, productivity, and system reliability.
An MCP server that turns raw meeting transcripts into summaries, correctly-owned action items, tasks, and follow-ups — as one autonomous agentic chain, not a manual checklist.
Artha turns any MCP client into a finance & tax copilot for Indian taxpayers — running on live, keyless data (real AMFI NAVs, Razorpay bank checks, Finance Act 2025 slabs). Its deterministic advisory council debates invest-vs-prepay with zero extra LLM calls and achieving nearly zero hallucinated numbers.
The Autonomous Product Intelligence Platform (ADIP) is an enterprise-grade, microservice-based AI pipeline that autonomously:
🔍 Explores any web application via headless browser automation
🧠 Constructs a structured knowledge graph of entities, relationships, and UI flows
📄 Generates comprehensive user guides, admin documentation, and FAQs
🧪 Runs automated smoke tests, regression tests, and produces test reports
🎬 Produces professional demo videos with 3D WebGL animations, AI voiceover, and synced subtitles
📦 Tracks version changes and auto-generates release notes and change summaries
The monolithic agent system has been decomposed into 7 independent MCP microservices governed by a centralized FastAPI Orchestrator, enabling horizontal scaling, overload protection, and fault isolation.
Work Sight AI is an intelligent workplace monitoring system that uses AI-powered computer vision to automatically track employee attendance, detect phone usage, and monitor break times in real-time. The system consists of a camera-based detection service that identifies faces and phones using YOLO AI, a backend API that stores all data in a database, and a live dashboard that displays everything from attendance numbers to focus scores. Managers can view the live status of their workplace, see who is present, receive instant alerts when someone uses a phone, and generate automated reports—all through a user-friendly interface. Built as a 24-hour hackathon project, Work Sight AI transforms manual workplace monitoring into a fully automated, data-driven solution that helps organizations improve productivity and employee engagement.
Memora is an AI-powered study companion designed to make learning more effective and engaging. Students can upload PDFs, lecture notes and the platform automatically converts them into concise summaries, quizzes, flashcards, and concept-based explanations.
Using NitroStack MCP, Memora provides personalized learning recommendations, tracks progress, and identifies weak areas to help students focus on what matters most. It transforms traditional study materials into an intelligent, interactive learning experience that improves understanding and long-term retention.
Ekalavya is an AI-powered career intelligence platform built as a native MCP server that fundamentally reimagines how people grow into their careers — not through passive tutorials or rote memorization, but through real projects, live collaboration, and ruthlessly personalized AI guidance. Upload your resume and Ekalavya's AI instantly dissects your skills, surfaces your blind spots, and maps the exact gap between where you are and where you want to be; paste any job description and get a precision-built learning roadmap tailored to close those gaps fast. The Group Foundry lets you create or join live collaborative workspaces where peers build real projects together using session codes, turning solo grinding into team momentum. An always-on AI career coach remembers your goals, tracks your growth stage from Seed to Tree, and gives you brutally honest, context-aware guidance every step of the way. Community channels keep you plugged into your domain's pulse, while intelligent job matching ranks live opportunities by how close your current skills are to the role — so you stop guessing and start targeting. Being MCP-native means Ekalavya plugs directly into Claude, ChatGPT, or any AI agent, turning your existing AI assistant into your personal career accelerator — no new app to learn, no friction, just results.
> QuickER is an AI-powered emergency response platform designed to save lives by reducing the delay between an accident and medical assistance. In emergency situations, users can instantly report incidents through text, voice, or images. Our AI analyzes the information, determines the severity of the emergency, extracts the location, and identifies the nearest hospitals, ambulances, police stations, and other emergency services. The platform intelligently prioritizes cases and enables faster coordination between victims and responders.
QuickER also provides step-by-step first-aid guidance while professional help is on the way, allowing bystanders to take immediate action during the critical golden hour. The system keeps users informed with real-time updates on emergency requests and responder status.
Built using AI, geolocation, intelligent automation, and modern backend technologies, QuickER aims to make emergency response faster, smarter, and more reliable. It reduces manual effort, improves communication between emergency services, and ensures that the right help reaches the right place at the right time.
Our vision is to create a unified emergency assistance platform that is accessible to everyone, minimizes response time, and increases the chances of survival during medical emergencies, road accidents, natural disasters, and other critical situations. By combining intelligent decision-making with seamless coordination, QuickER transforms emergency response into a faster, more efficient, and life-saving experience.
WHAT DOES IT DO?
ForgeMind AI is an Autonomous AI Factory Operations Copilot built for smart manufacturing (Track 02). It continuously monitors Industrial IoT sensor streams (temperature, vibration, pressure, RPM, current) across factory machines. When an anomaly occurs, ForgeMind AI predicts failure risk (98% precision, ~4h Remaining Useful Life), performs Explainable AI root-cause analysis, verifies spare parts in inventory, simulates rerouting production load to backup machines with zero downtime, and auto-dispatches emergency maintenance tickets (TKT-1001).
WHO IS IT FOR?
It is designed for Plant Managers, Maintenance Engineers, Reliability Supervisors, and Operations Teams in manufacturing industries such as automotive assembly, textile mills, CNC machining, and heavy manufacturing. It transforms raw sensor data into coordinated, instant operational decisions.
WHAT MAKES IT SPECIAL?
1. 7-Agent MCP Architecture: Instead of an isolated chatbot, ForgeMind AI deploys 7 specialized agents and 14 MCP tools working together in an end-to-end operational chain.
2. What-If Production Reroute Simulation: Rather than just warning about a broken machine, it evaluates entire factory capacity and simulates moving work to backup units, saving ₹12 Lakhs to ₹75 Lakhs in downtime loss per event.
3. Explainable AI Root Cause Analysis: It correlates multi-sensor trends (Temp +36.7%, Vibration +23.2%) with 92% confidence instead of giving black-box answers.
4. Interactive Live UI Dashboard: Features a custom cyberpunk React widget with live 10-second auto-fluctuating telemetry stream.
5. 24/7 Autonomous Background Sentinel: Runs continuously on NitroStack, auto-dispatching alerts even when supervisors are offline.
Genesis AI is an AI platform that turns any idea into a complete execution plan. Instead of using one AI, it uses multiple AI agents like Planner, Finance, Marketing, and Risk experts to work together and generate a business plan, budget, timeline, and strategy in minutes.
Most "AI financial advisors" are a chatbot bolted onto a bank app: you ask a question, an LLM answers from whatever context it's given, and the moment you close the tab, it stops thinking about you. We Win is built to be structurally different — a system that watches, calculates, and acts on a user's behalf continuously, while the AI itself is never trusted to touch a single number.
Deterministic core, not a chatbot guessing. Every score you see — the six-dimension Financial Health Score, the Safe-to-Spend limit, the Credit Trust Score — is computed by hard-coded, testable Python engines benchmarked against RBI, SEBI, and IT Act standards. The LLM layer (Google Gemini, orchestrated via LangChain and CrewAI) is architecturally forbidden from performing arithmetic; it only ever narrates numbers that have already been verified. This isn't a prompting choice — it's a hard boundary, which means the system cannot hallucinate a financial figure by construction, not by hoping the prompt holds.
A crew of agents, not one model doing everything. Instead of a single LLM handling every task, We Win runs six narrowly-scoped agents — Aggregator, Context & Incident, Market Watch, Loan Matching, Explainer, and Execution — each with exactly one job and exactly the MCP tools needed to do it. None of them ever holds a database credential. Every read and every write is mediated through a purpose-built MCP tool, so a hallucinated or malformed request from an agent can never reach Supabase or Neo4j directly — a non-negotiable requirement when the data is financial.
Genuinely proactive, not reactive. A static balance tells you what you have; it doesn't tell you what you can safely spend after tomorrow's EMI and next week's rent. We Win's Safe-to-Spend Guardian nets obligations against income in real time and alerts users before they overdraw — while a life event (a new baby, a job change) instantly re-weights only the affected part of their plan, without re-running the whole system. Not
LOOM is an AI-assisted footwear compatibility platform designed to solve one of the biggest challenges in footwear shopping: finding shoes that truly fit. While traditional recommendations rely almost entirely on shoe size, real comfort depends on multiple factors, including foot geometry, movement patterns, lifestyle, intended activity, and individual needs. Two people wearing the same shoe size can have completely different experiences with the same pair of shoes.
Med-AgenticAI MCP is an AI agent that sits on top of a centralized patient record system and turns it into real-time safety and coordination: when a hospital logs a visit, it automatically cross-checks the new prescription against that patient's entire history across every hospital they've ever been to — flagging drug conflicts, duplicate tests, and allergies no single facility's records could reveal — reroutes prescriptions to a sister facility or triggers replenishment if a medicine is out of stock locally, and recommends a follow-up timeline with a stated medical reason based on diagnosis severity. It's built for doctors, who gain visibility they'd never have from their own hospital's records alone; for patients, who get safer prescriptions and continuity of care; and for hospitals and pharmacies, who gain network-level coordination instead of operating in isolation. What makes it special is that its core capability — catching cross-hospital conflicts — is only possible *because* the records are centralized, and every flag it raises is traceable to the exact hospital, date, and record that caused it, making it a genuine demonstration of centralized data enabling reasoning that was previously impossible, not just automation dressed up as AI.
AI Recipe & Nutrition Assistant is an intelligent AI-powered application built using the NitroStack MCP (Model Context Protocol) Framework. It enables users to discover recipes, explore ingredients, and analyze nutritional information through natural language conversations.
The system integrates two comprehensive datasets—a recipe dataset and a nutrition dataset—and exposes them as MCP tools that AI agents can access in real time. Instead of manually searching through multiple websites or nutrition databases, users can simply ask questions such as "Find a high-protein breakfast recipe", "Show foods rich in fiber", or "How many calories are in an apple?". The AI agent interprets the request, invokes the appropriate MCP tools, retrieves relevant information from the datasets, and delivers accurate, structured, and easy-to-understand responses.
Ask any AI which government savings schemes you qualify for and it answers confidently — from memory. We tested it: on edge cases it was wrong in 14 of 30 runs, and contradicted itself 5 times on identical questions.
FinBridge AI is an MCP server that replaces the guess with a lookup. It evaluates all 7 major Indian schemes — PMJDY, APY, PMJJBY, PMSBY, SSY, SCSS, NPS — against a rulebook sourced from government circulars, and names the exact condition that failed for every rejection. It also projects mutual fund growth from live NAV data as an honest range, scores financial health, and explains financial terms in plain language.
Across the same 30 test runs, FinBridge got zero wrong.
The model explains. The code computes.
What it does: Given one goal, FinPilot autonomously ingests transactions, analyzes spending, detects risks, and delivers a complete recommendation — no manual steps.
Who it's for: Students juggling hostel fees, internship income, and irregular budgets.
What makes it special: It's agentic, not conversational — it acts, not just answers. It reads spending behavior, not just numbers, and lets you simulate real decisions ("what if I move to Bangalore?") before you make them.
MCP server that turns a photo + measurements of a any bottle cap or bin lid into a 3D-printable replacement — precise threaded caps and hole-fill plugs via CadQuery, plus image-to-mesh via Meshy, with PyMeshLab-based mesh repair. Built on NitroStack.
FactoryOS is a next-generation, autonomous Smart Manufacturing Control Center engineered for Industry 4.0. Built on top of the Model Context Protocol (MCP) and the NitroStack framework, FactoryOS connects real-time Internet of Things (IoT) physical floor telemetry to a collaborative, multi-agent AI decision-making engine.
Traditional manufacturing plants operate in reactive survival mode: equipment breaks down unexpectedly, maintenance teams scramble to locate spare parts, plant managers manually make phone calls to suppliers to negotiate delivery lead times, and assembly lines sit idle for hours or days. This reactive paradigm costs global industry over $500 billion annually in unmanaged downtime, lost productivity, and supply chain inefficiencies.
FactoryOS fundamentally transforms factory operations from reactive firefighting to proactive, self-healing autonomy. By continuously monitoring machine sensor metrics (temperature, vibration, rotational speed, torque, process temperature, and tool wear), FactoryOS utilizes a Scikit-Learn Random Forest Machine Learning model to predict hardware component failures days before they occur.
When a risk or anomaly is detected, FactoryOS deploys a Supervisor Agent that orchestrates five domain-specialist AI agents (Maintenance, Inventory, Procurement, Production, and Safety). Operating over type-safe MCP tools, these agents autonomously coordinate to evaluate failure probabilities, verify safety stock thresholds, negotiate expedited component orders with external vendor networks, reroute manufacturing job schedules to alternative assembly lines, and compile OSHA-compliant safety audit trails.
By automating the complex multi-step triage process that previously required hours of manual human intervention, FactoryOS achieves up to a 92% reduction in downtime financial losses, preserving thousands of dollars per operational incident while guaranteeing strict worker safety and regulatory compliance.
NexusOS is an AI-powered Enterprise Incident Response Platform that orchestrates specialized MCP agents for Security, Compliance, Communication, and Decision Making. It provides a unified Crisis Command Center inside NitroStack Chat, enabling organizations to detect, coordinate, and resolve enterprise incidents faster through intelligent multi-agent collaboration.
Every year, millions of citizens—especially students, low-income families, and rural applicants—miss out on vital government scholarships, welfare schemes, and financial aid simply because the system is too complicated. Between navigating dense eligibility rulebooks, figuring out which documents are actually required, and dealing with confusing government portals, the application process is an obstacle course. Furthermore, people are rightfully hesitant to upload sensitive financial documents and personal IDs to unverified third-party assistance tools.
We are building GuardianAI, an intelligent, privacy-preserving digital caseworker powered by the Model Context Protocol (MCP). Think of it as a personalized advisor that guides applicants from initial eligibility verification all the way to portal submission, without ever compromising their data security.
Instead of forcing users to fill out endless forms, GuardianAI uses conversational intake to build a "minimized profile." Our built-in Policy Engine ensures that only non-sensitive data (like age, state, or general student status) is used to query eligibility rules, while highly sensitive details (like Aadhaar, PAN, or bank account numbers) are tokenized and masked. Once a user is verified for a scheme, GuardianAI generates a custom document checklist, validates their uploaded certificates, and automatically queues the application for submission on government portals.
Why this matters: Financial inclusion isn't just about opening bank accounts; it’s about ensuring that government capital and educational subsidies actually reach the people they were designed to help. By combining conversational AI with strict privacy guardrails and automated workflow planning, GuardianAI removes the bureaucracy, language barriers, and data-privacy fears that keep citizens from claiming their rightful financial benefits.
ThreatMatrix is an AI-powered cyber threat investigation platform that enables users to instantly analyze suspicious URLs, files, emails, QR codes, and digital content through a single intelligent interface. Powered by the Model Context Protocol (MCP), ThreatMatrix connects with multiple cybersecurity tools and threat intelligence services to automate evidence collection, risk assessment, and security analysis.
Designed for students, internet users, developers, IT teams, and cybersecurity professionals, ThreatMatrix eliminates the need to manually verify digital threats across multiple platforms. Instead, it provides a comprehensive investigation by assigning a risk score, identifying potential indicators of compromise, explaining the reasoning behind its analysis, and recommending appropriate actions.
What makes ThreatMatrix unique is its ability to orchestrate multiple security tools through MCP while leveraging AI to correlate findings into a single, explainable report. Rather than simply labeling content as safe or malicious, ThreatMatrix delivers transparent, evidence-based insights, helping users make informed security decisions quickly and confidently.
"An AI council that debates, verifies, and recommends the best path forward."
Synod AI is a multi-agent decision intelligence system that helps engineering teams make reliable, evidence-backed decisions through autonomous AI collaboration.
Modern technical decisions often involve uncertainty, conflicting opinions, and incomplete information. Synod addresses this by creating an AI-powered decision council where specialized agents analyze a problem from multiple perspectives before producing a final recommendation.
The system combines multi-agent reasoning, Retrieval-Augmented Generation (RAG), and evidence verification to simulate a structured decision-making process.
A decision submitted by the user goes through multiple stages:
Context Intelligence: Understands the decision, objectives, constraints, assumptions, and identifies missing information.
Advocate Agent: Builds the strongest case in favor of the proposed decision by analyzing benefits, opportunities, and technical advantages.
Critic Agent: Challenges the decision by identifying risks, limitations, hidden assumptions, and possible failure scenarios.
Evidence Intelligence: Uses RAG-based retrieval and verification to evaluate supporting information and measure evidence confidence.
Verdict Agent: Synthesizes all perspectives and generates a final recommendation with reasoning, confidence score, risks, and next steps.
Built as an MCP-native AI application, Synod provides structured tools, resources, and interactive workflows that allow AI systems and users to seamlessly interact with the decision engine.
Unlike traditional AI assistants that provide direct answers, Synod creates a transparent reasoning process where decisions are debated, validated, and justified before recommendation.
Synod AI transforms AI from a simple answer generator into an intelligent decision council.
GeoTrust is an all-in-one compliance platform built to simplify and secure your location-based operations. We provide a single, unified solution for managing geographical data verification, identity authentication, and strict regulatory adherence. By consolidating complex security protocols into one accessible system, GeoTrust eliminates the need for multiple third-party tools. Our platform guarantees data integrity, offers complete transparency, and ensures that your business remains fully compliant with industry standards, all while operating seamlessly from a centralized dashboard.
ALE Supply Chain Management (SCM) is an AI-native backend server built on NitroStack that connects enterprise ERP workflows directly to conversational AI clients via the Model Context Protocol (MCP). By combining open-source LLM document OCR, distributed SQLite Cloud storage, and interactive Next.js chat widgets, the system autonomously ingests unstructured invoices, performs automated 3-way matching against Purchase Orders with strict tolerance rules, and seamlessly escalates discrepancies to human finance teams through real-time email alerts.
FactoryWise AI is an Agentic AI-powered decision support platform that helps manufacturers make smarter modernization decisions for legacy factory machines. Instead of simply predicting failures, it analyzes machine health, maintenance history, energy consumption, operational performance, and financial impact to recommend whether a machine should be Continued, Retrofitted, Upgraded, or Replaced.
The platform uses multiple specialized AI agents orchestrated through NitroStack MCP, where each agent independently gathers data, performs domain-specific analysis, and collaborates to generate an explainable, ROI-driven recommendation. Every decision includes business reasoning, confidence scores, estimated ROI, and actionable next steps, enabling factory managers to reduce costs, minimize downtime, and accelerate Industry 4.0 adoption through intelligent, data-driven modernization strategies.
ContextOS is a permanent, queryable engineering memory layer that sits on top of your team's existing tools (GitHub, Slack, codebase) and answers technical questions with full citations — no hallucinations.
Engineering knowledge lives in scattered Slack messages, old PRs, and senior engineers' heads. When someone new joins and asks "Why did we choose Redis?" or "What breaks if I change CheckoutService?" — nobody knows where to look.
Its meant for Engineering teams. Specifically: new engineers who need to get up to speed fast, senior engineers who are tired of answering the same questions repeatedly, and engineering managers who want institutional knowledge to survive team turnover. Any company using GitHub and Slack can plug it in immediately — no migration, no setup.
Crawls through the public channels of telegram which are being used for betting promotions and operations. It finds other possible channels from the initial channels, creating a snowball effect and forming a network of betting channels to be reported
An MCP-powered decision engine built with NitroStack to automate loan processing, policy validation, and real-time financial risk assessment through interactive UI widgets.
Citizens report civic issues (potholes, leaks, garbage, streetlights) with a photo, voice note, or GPS pin, and instead of it disappearing into a queue, a real AI agent takes over: it plans the steps, calls actual tools to classify, assess severity, route it, and set a deadline — then either finishes the job on its own or stops and asks a human to approve. It's built for residents (simple reporting + real tracking), officials (routine work pre-triaged, only pulled in when needed), and the institution (full audit trail of every decision).
What makes it special: it's a genuine agent, not an AI label-generator — you watch it plan and act live through real MCP tool calls, not a canned result. And it knows when not to act alone — a visible approval gate blocks anything risky until a human says yes. That's what makes it trustworthy enough for government use, and portable enough to work anywhere an AI agent needs to be trusted with real actions.
Built on the Model Context Protocol (MCP), three LLM agents — Scout, Analyst, and Skeptic — debate each other through shared resources to produce a signal you can actually interrogate. A Next.js widget renders the entire handoff: headline found → signal scored → skeptic challenge → final verdict.
WARDEN is an MCP server for security teams drowning in vulnerabilities — thousands open, time to fix five.
It does three things nobody combines. It ranks by evidence, not theory: instead of a severity score that says what could go wrong, it checks whether a bug is actually being exploited (CISA KEV) and how likely it is to be hit (FIRST EPSS) — so a "terrifying" 9.8 nobody attacks is deprioritised, and a modest 7.5 used by ransomware crews gets patched today.
It defends itself. Threat-intel work means feeding AI attacker-written text — the perfect vehicle for prompt injection. WARDEN never hands raw fetched pages to the model: it extracts facts server-side, quarantines hidden commands, and flags the attempt without being affected by it.
And it knows its limits. It auto-fixes what it safely can (a ready-to-review patch PR), flags what needs a human, and refers the rest — but it never touches production, by design.
Every decision, including rejected leads, is fully auditable. Open Innovation / cybersecurity.
An organizational knowledge platform that turns scattered documents, repositories, and
engineering activity into a queryable memory, and exposes that memory to AI clients over the
Model Context Protocol (MCP).
Built for the NitroStack x Amrita University Hackathon.
ScholarPilot is an AI-powered Research Operating System built using NitroStack and the Model Context Protocol (MCP). It transforms fragmented academic research into a seamless, end-to-end workflow by coordinating specialized AI agents throughout the research lifecycle.
Starting from a research topic, ScholarPilot discovers and ranks relevant literature, extracts key claims, identifies research gaps, performs adversarial reviews to challenge weak ideas, evaluates research resilience, assists with writing, verifies citations and technical accuracy, preserves research memory, and exports manuscripts to Overleaf.
Unlike conventional AI writing assistants, ScholarPilot functions as a collaborative AI research committee, where multiple specialized agents analyze, critique, and refine research before publication. This results in higher-quality, evidence-backed, and more reliable academic outputs while significantly reducing the time required for literature review and manuscript preparation.
Built with NitroStack and MCP, ScholarPilot demonstrates how interoperable AI agents can revolutionize academic research by making the entire research process faster, smarter, and more trustworthy.
AEIOS-X is an MCP-native Enterprise AI Operating System that enables intelligent enterprise automation through dynamic AI orchestration. Instead of relying on a single AI assistant, it coordinates specialized agents using an execution pipeline with intent detection, knowledge management, shared blackboard memory, creates specialized agents with our agent factory instead of developer creation and consensus-based decision making. Built on the Model Context Protocol (MCP), AEIOS-X exposes enterprise capabilities through Resources, Tools, Prompts, and Tasks, enabling seamless interoperability with MCP-compatible applications. The platform features a real-time Enterprise Command Center, modular FastAPI backend, and scalable architecture designed for collaborative AI, enterprise decision support, and intelligent workflow automation across multiple industries.
Autonomous Digital Arrest & Mule-Account Interceptor A Model Context Protocol (MCP) Infrastructure Layer for Real-Time Digital Public Safety & Fraud Prevention
Investigation Context Platform is a cybersecurity investigation management system built on the NitroStack MCP Framework. It enables AI agents and security analysts to create, manage, and query investigation data through standardized MCP tools, resources, and prompts. The platform maintains persistent investigation context using JSON-based storage, allowing analysts to track incidents, manage evidence, correlate entities, record notes and tasks, generate reports, and perform contextual searches across ongoing investigations.
AgentSentinel is an AI-powered security platform that helps organizations discover, monitor, and secure AI agents operating across enterprise environments. As businesses increasingly adopt AI assistants, autonomous workflows, and AI-powered applications, many agents operate without proper visibility or governance, creating “Shadow AI” risks.
AgentSentinel continuously scans enterprise systems to discover AI agents, builds a live inventory, evaluates their security posture, and assigns risk scores based on permissions, data access, behavior, and compliance policies. The platform detects suspicious activities such as unauthorized data access, excessive permissions, and abnormal agent behavior.
When a high-risk agent is identified, AgentSentinel can automatically take protective actions such as quarantining the agent, revoking sensitive permissions, or notifying security teams. It also provides explainable AI-driven recommendations that help administrators understand why an agent was flagged and how to remediate the issue.
Built using the Model Context Protocol (MCP) and NitroStack, AgentSentinel exposes security capabilities as MCP tools and resources, allowing enterprise AI systems to integrate seamlessly with existing workflows and security operations.
Copilot finishes your code. MENTOR makes you finish it.
A student picks a real project and a role on it, gets the slice they would
actually own in a company, designs that slice on a canvas, then builds against
checkpoints derived from their own design. When it breaks, MENTOR compares what
they planned against what they built and names the decision that caused it — a
file and a line — and then refuses to write the fix.
The refusal is the product. The concept flashcard is released only after the
student's own tests pass, and only one of the three services ever holds an
answer, so a bug in the other two cannot leak what the student is meant to earn.
23 tools across three deployed MCP services (8 · 6 · 9), one loop.
The usage of the MCP was also implemented at https://github.com/nitrostacklh/command-global.git
An intelligent, multi-agent system that uses LLMs to autonomously manage, monitor, optimize, and secure Azure cloud infrastructure. Built with a dual architecture — a React web dashboard for end-users and an MCP server for AI-native tool orchestration via NitroStack, Claude Desktop, or any MCP-compatible client.
Zero-Trust-Commander is an autonomous infrastructure remediation gateway built on the NitroStack framework. It empowers AI agents to investigate production outages, trace broken commits, and analyze system health—while strictly enforcing a Human-in-the-Loop Zero-Trust Security Gate for all destructive or mutating actions.
Seer enables non-technical users to ask predictive questions about datasets using natural language. It profiles the dataset, identifies data-quality issues, and prepares a transparent regression or classification plan for the user to review and approve. Once approved, Seer automatically handles missing values, encoding, scaling, model training and evaluation against a simple baseline. The final result is presented through interactive widgets containing predictions, metrics, graphs, warnings and plain-language limitations.
Problem: Ask a Swiggy or Zomato assistant what to eat tonight and it answers from menus, ratings and past orders. It cannot see your HbA1c, your ferritin, or that you slept four hours. It cannot know you're on warfarin, so a spinach curry may interfere with your medication. These platforms optimise for conversion, personalisation stops at cuisine preference. Fitness trackers hold the other half of the picture (sleep, stress, recovery, activity) in dashboards that never reach the ordering decision.
Solution: NutriPulse is an MCP server unifying clinical profiles, lab panels, live wearable telemetry and a USDA-resolved food catalogue into one layer any AI agent can reason over.
Real-time biometrics drive it. It accepts sleep, stress, HR recovery, hydration, steps and active calories from a smartwatch or band, and targets shift immediately, four hours' sleep with stress at 85 raises protein, lowers the sugar ceiling and increases fluid requirements, each adjustment naming its trigger.
Recommendations run a four-axis resolver. Every candidate is scored independently on clinical fit, contextual/taste fit, budget fit and craving satisfaction. The engine computes the Pareto-nondominated set, then applies a lexicographic tiebreak: clinical severity, then budget cap, then craving, then preference. Allergen and drug–nutrient interaction BLOCKs are absolute, enforced by a framework-level guard no tool can bypass. Missing nutrient data fails closed rather than silently passing.
Every result returns a conflict log naming what was sacrificed in real units: milligrams, grams, rupees and which rule decided it. Computed, not generated.
Impact: India carries one of the world's largest diabetic and hypertensive populations alongside surging delivery adoption. NutriPulse negotiates healthier swaps rather than refusing cravings.
External components: NitroStack SDK; USDA FoodData Central (traceable source IDs); Open-Meteo. Synthetic personas; no real patient data.
APIGuard is an API release-safety system. It compares OpenAPI contracts, finds breaking changes, traces their impact across consumer repositories, uses an LLM only where evidence is ambiguous, applies release policy, records an auditable human decision, and creates guarded draft migration pull requests.
It is essential because an API provider’s tests can pass while downstream applications still break. APIGuard turns “this schema changed” into verified evidence showing exactly which repositories, files, and owners are affected—and provides a controlled path to fix them.
AuditPulse is an autonomous, multi-agent AI system that unifies enterprise auditing. It dynamically analyzes HR, Finance, and Security databases to detect overbilling, compliance leaks, and unusual activity, instantly generating interactive dashboards and executing secure, PII-masked workflows.
Modern employees receive hundreds of notifications every day from workplace tools such as Slack, Microsoft Teams, Email, Jira, GitHub, Calendar, and PagerDuty. Since every notification appears equally important, employees spend valuable time deciding what to read first. Critical updates often get buried among routine messages, leading to missed deadlines, delayed responses, and reduced productivity. While existing AI tools can summarize notifications, they do not understand the business context or identify which notifications require immediate attention.
ImmunoGraph is a vaccine and therapeutics research workflow, a multi-agent-driven platform built with MCP. It helps researchers analyze protein sequences, generate candidate peptides, predict immune binding behavior, rank candidates, examine supporting evidence, and export structured research outputs.
It is aimed at scientists, bioinformatics teams, and AI agents that need reproducible immunology workflows, via a Model Context Protocol server. It’s special in that it combines deterministic scientific tooling, provenance-aware outputs, graph-style workflow modelling and MCP integration so that the same backend can serve both human-facing apps and agent-driven research automation.
An MCP based orchestration server that enables AI agents to analyze repositories, detect conflicts, manage project knowledge, and generate documentation, streamlining software engineering workflows with automated, context aware collaboration.
Agentic AI CCTV Analyzer is an intelligent surveillance platform that transforms traditional CCTV systems into AI-powered security assistants. Instead of requiring security personnel to manually review hours of video footage, the system automatically analyzes surveillance videos, detects significant events, organizes them into structured timelines, and enables users to retrieve critical information through intelligent search.
Cross-Functional Review Board (CFRB) is a multi-agent AI platform that simulates how enterprise proposals are reviewed by Security, Legal, Engineering and Product teams before implementation. Instead of waiting days for manual reviews, specialized AI agents analyze proposals in parallel, identify risks, detect policy violations, estimate costs, and generate actionable recommendations. A moderator agent consolidates the results into a final decision with supporting evidence and an approval score. The platform provides explainable reasoning, traceable workflows, and enterprise-ready governance, helping organizations make faster, more informed, and compliant decisions.
Most delivery feedback is subjective — "sound more confident," "be more engaging" — advice that isn't actionable and, more importantly, isn't something an agent can honestly claim to fix. Confidence is built by talking to real people, not software. What is fixable and checkable: how many filler words you used and where, whether you rushed or dragged relative to your own baseline pace, and whether your vocal stress landed on your script's actual key points or drifted off them. This project deliberately narrows the scope from "make you a better speaker" to "show you precisely where your delivery didn't match your plan" — a claim it can actually back up with timestamps and numbers.
4ex is an Enterprise Knowledge Integrity MCP Server built on the NitroStack framework. It acts as an intelligent, automated auditor that continuously monitors corporate knowledge bases, detecting conflicts between live operational data and authoritative company policies (such as unstructured PDFs). When a contradiction occurs, 4ex instantly identifies the policy drift, calculates the associated risk, and generates an automated remediation report. This tool is designed for modern enterprise compliance, security, and IT teams who struggle with maintaining data alignment across siloed departments and currently spend days manually auditing policies. What makes 4ex special is its Hybrid Data Architecture, which goes beyond simple API wrappers; it synchronously parses unstructured binary documents (PDFs) and merges them with structured JSON systems into a unified, single source of truth, empowering AI agents to proactively and autonomously secure corporate knowledge in seconds.
An MCP server that enables AI agents like ChatGPT, Claude, and Gemini to automate GitHub workflows and cloud deployments. It analyzes repositories, detects the tech stack, generates Dockerfiles and test-gated CI pipelines, creates commits and pull requests, performs deployment readiness and security checks, and deploys applications to Vercel, Netlify, Cloudflare Pages, and Render—all through natural language conversations.
SUMO AI MCP is an intelligent urban traffic simulation and analytics platform. It connects Eclipse SUMO (Simulation of Urban Mobility) with the Model Context Protocol (MCP) via NitroStack (TypeScript) and FastMCP (Python).
Instead of manually crafting complex XML files or executing multi-step CLI commands, users and LLM agents can design, run, visualize, and audit microscopic traffic simulations through natural language.
n AI agent for EOD-driven team intelligence. Employees write one honest paragraph a day; the agent verifies it against what actually happened in GitHub, and tells managers only what genuinely needs them.
AI agents can read financial data, draft reports, and process invoices — but no finance team will let one move money unsupervised. The blocker is trust: an agent follows instructions, and instructions can be manipulated. AegisPay solves this by moving enforcement out of the prompt and into the server. Every payment attempt passes through a deterministic risk engine, a role-separated approval gate, and a tamper-evident audit trail — none of which the agent can argue with, override, or bypass through prompt injection. Built entirely on the NitroStack TypeScript SDK with 58 passing unit tests on the risk engine.
WHAT DOES IT DO?
IT Access Resolver converts complex, manual Level-1 IT diagnostic workflows into an autonomous, self-healing orchestration engine using the Model Context Protocol (MCP). When an employee experiences a SaaS access failure, license seat exhaustion, account lockout, or VPN networking blocker, our system continuously performs multi-tier diagnoses and automated remediation—provisioning directory group memberships, expanding software license pools, and resetting Zero-Trust network handshakes without human intervention.
WHO IS IT FOR?
1. Enterprise IT Operations & Security Teams: Frees support agents from repetitive L1 ticket triage, reducing operational overhead, MTTR (Mean Time to Resolution), and manual log correlation across isolated identity and networking silos.
2. Employees: Provides an instantaneous self-service portal within AI client interfaces where they can report access blockers and have their permissions restored in real time, eliminating lost productivity while waiting for manual ticket assignees.
3. IT Governance & Compliance Auditors: Automates software seat reclamation from inactive users and ensures SOC-2 compliance by tracking every automated remediation in an immutable, event-driven audit trail.
WHAT MAKES IT SPECIAL?
• Zero-Credit Employee Self-Service UI: Features custom interactive micro-frontend React widgets (@Widget hooks) embedded directly inside the AI chat interface. Employees can view real-time diagnostic steppers, filter tickets, and trigger fixes via direct tool calls (callTool) at $0 in LLM computation or token credits.
• End-to-End Autonomous Orchestration: Seamlessly couples Standard Operating Procedure (SOP) prompt workflows with real-time JSON Resource streams (identities, licenses, network telemetry) and atomic tools to run multi-step diagnostic decision trees.
• SOC-2 Event-Driven Architecture & Governance: Built on a modern NestJS-style dependency injection infrastructure featuring strict Zod validation, built-in
AI-powered conversational clinical deterioration explainer. Our goal is to help bedside nurses and doctors understand why a patient is at risk, instead of only showing a static risk score.
Curio is an AI-powered gamified learning engine built on the Model Context Protocol (MCP) using NitroStack. It transforms any educational topic—from Photosynthesis and Quantum Physics to History, Law, and Medicine—into a fully interactive, voice-narrated 2D platformer game in real time with a single AI prompt.
WHAT IT DOES: Curio instantly generates custom 3-level platformer games for any concept. Each game includes escalating difficulty, interactive checkpoint puzzles (Sequencing, Collecting, and Matching), end-of-level boss quizzes, dynamic visual themes, and an AI voice narrator. One simple prompt generates a complete, playable educational world in seconds.
WHO IT IS FOR: Curio is designed for students, educators, self-learners, and schools looking for an active alternative to passive textbooks and video lectures. It enables learners of any age to experience and play through concepts across any subject or field of knowledge.
WHAT MAKES IT SPECIAL:
Active Memory Recall: When a checkpoint challenge starts, the lesson text completely disappears, forcing learners to retrieve knowledge from memory—scientifically proven to maximize retention and overcome the Forgetting Curve.
Physical Mastery Gates: Laser energy barriers physically block forward progress until the learner solves the checkpoint puzzle, guaranteeing mastery before advancement.
Human Prosody Voice Engine: Custom clause-level chunking and pitch modulation create natural, human-like teaching narration instead of robotic text-to-speech.
Zero-Dependency Synthesized Audio: All 8-bit sound effects are generated mathematically at runtime using the Web Audio API, requiring zero external audio file downloads.
Universal MCP Architecture: Built on the open MCP standard, any AI agent (Claude, GPT, Gemini) can call Curio's tool to publish interactive learning games instantly.
Curio collapses weeks of traditional instructional design into a 3-second AI prompt, turning human curiosity into unforgettable learning exp
The Industrial Edge Machinery & Workload Health Monitoring System is an AI-driven, real-time telemetry and diagnostic platform engineered for smart manufacturing, industrial automation controllers (IPCs), edge computing nodes, and factory robotics. Built on the Model Context Protocol (MCP) using FastMCP and LangChain, the system enables autonomous AI agents and maintenance engineers to continuously monitor hardware compute health across CPU, RAM, and NVIDIA GPUs (auditing per-core load, thermal zones, VRAM, and power draw to prevent overheating or compute bottlenecks), inspect low-level Linux kernel ring buffers (`dmesg`) and OS system logs for hardware driver failures or Out-of-Memory crashes, and audit software logs from edge vision models and SCADA gateway processes. Upon detecting critical equipment anomalies or process failures, the platform automatically dispatches structured, root-cause incident reports via email to plant reliability engineers for proactive, predictive maintenance.
Disaster Response MCP is a Model Context Protocol (MCP) server that provides AI assistants with intelligent disaster response and emergency coordination capabilities. It supports multilingual voice and text emergency reporting, automatically analyzes incidents, creates and tracks SOS requests, recommends nearby emergency shelters, provides real-time area risk assessments, enables human-in-the-loop decision making, and monitors live incident status and location updates through standardized MCP tools, enabling faster, more accessible, and coordinated emergency response during natural and man-made disasters.
FlowLogix is a multi-agent MCP server that turns an LLM into an autonomous warehouse operations co-pilot.
It exposes all six fulfillment stages — inbound receiving, putaway, cold-chain monitoring, order picking, emergency purchasing, and vendor dispatch — as AI-callable tools via the Model Context Protocol. Critical business logic runs deterministically on a NitroStack backend (no hallucinated math), while remote Slack and Mail MCP servers handle real-world alerts and vendor emails over SSE. High-impact actions pause for human sign-off via React widgets before execution.
MultiverseOps is an enterprise-grade agentic control platform engineered to eliminate operational risk during high-stakes IT, DevOps, and governance deployments. Built on the Model Context Protocol (MCP) standard and powered by Google Gemini AI, the platform speculatively evaluates proposed operational workflows across thirty concurrent virtual environments—stress-testing each action against infrastructure bottlenecks, cybersecurity threats, data schema drift, and regulatory compliance constraints in sub-two-second latency. By analyzing real-time simulation telemetry, MultiverseOps autonomously synthesizes remediated execution plans and safely dispatches validated updates to production systems, providing complete operational transparency through cryptographically signed SHA-256 audit trails.
DTaaS (Digital Twin as a Service) is an AI-powered MCP server that enables users to create complete IoT Digital Twin infrastructures using simple natural language prompts. Instead of manually configuring devices, dashboards, rule chains, alarms, users, and customers, users simply describe their requirements, and DTaaS automatically provisions the entire digital twin on Digital twin platforms.
The platform combines Large Language Models with deterministic infrastructure provisioning. An AI Planner interprets the user's intent and generates a structured Digital Twin specification, while an Engineer module executes the provisioning by interacting directly with the ThingsBoard APIs.
DTaaS supports the complete lifecycle of a digital twin by automatically creating:
IoT Devices
Dashboards
Smart Widgets
Rule Chains
Alarm Rules
Customers
Users
Device Emulators
Digital Twin Graphs
Manufacturing Waste-to-Revenue: An Agentic AI Multi-Agent System for Industry 4.0 . It is an agentic AI system— a network of autonomous AI agents, each running as an MCP server on NitroStack, that collectively apply Industry 4.0 priciple
Protocol-0 is an autonomous, multi-agent SRE copilot built on the NitroStack MCP framework. It continuously monitors enterprise infrastructure, diagnoses root causes, calculates business SLA impact, and executes auto-remediation plans, secured by a cryptographic Zero-Trust Execution Gate.
The Problem: Modern incident response is broken. SREs spend hours correlating alerts across fragmented systems (GitHub, Datadog, Jira, PagerDuty), while executives panic over business SLAs. Existing automated tools are either too simple or too dangerous, if an AI agent has write-access to your production servers, a single hallucination or prompt-injection could delete your databases.
The Solution: Protocol-0 solves this by wrapping the entire incident funnel into a deterministic, multi-agent LangGraph engine running natively over Model Context Protocol (MCP). It deploys 4 specialized agents:
Monitoring Agent: Polls Datadog and SaaS telemetry to detect anomalies.
Engineering Agent: Inspects Kubernetes pod failures and executes root-cause analysis.
Executive Agent: Projects business impact, SLA risks, and department health scores.
Action Agent: Implements remediation playbooks (scaling deployments, rolling back commits, extending deadlines).
The Secret Weapon: Zero-Trust HMAC Guard Security is enforced in code, not prompts. The Action Agent physically cannot mutate infrastructure without a valid cryptographic HMAC signature generated by the human-in-the-loop approval. Even if the AI is prompt-injected to "bypass safety," the zero-trust gate blocks unauthorized executions instantly.
Why It Matters: Protocol-0 transforms AI from a basic Q&A chatbot into a secure, collaborative, and autonomous SRE team member. It bridges the gap between raw infrastructure engineering and high-level executive decision-making, keeping your systems running and your production secure
Domain Expansion is an automated API Surface Security Analyzer built on the Model Context Protocol (MCP). It continuously scans API logs, traffic, and OpenAPI specifications to detect security gaps, authorization mismatches, shadow endpoints, data leakage, and log injection vulnerabilities.
Designed for modern DevOps and AppSec teams, it bridges the gap between raw telemetry and security remediation by serving structured findings, interactive topology graphs, and evidence viewers directly to AI agents. Domain Expansion transforms routine API security audits into an automated, real-time workflow that keeps enterprise applications secure by default.
Factory Mind, an AI-powered multi-agent decision platform for smart manufacturing. Instead of manually coordinating across different departments, the user simply enters a production-related query through our chat interface. The system then activates specialized AI agents such as Production, Inventory, Maintenance, Logistics, and Quality. Each agent analyzes the problem from its own perspective, and an orchestrator combines their insights to generate a single optimized decision. The platform not only identifies the root cause but also provides actionable recommendations, helping factory managers reduce downtime, improve efficiency, and make faster, data-driven decisions.
Project Aegis is a mathematical, closed-loop SRE control plane for FinTech that uses Agentic AI to prevent operational failure in high-concurrency Core Banking Systems. Traditional monitoring relies on static alerts that miss complex microservice degradation like database lock contention. Aegis replaces human response with sub-millisecond anomaly detection and an autonomous AI agent swarm, guaranteeing 99.999% availability and zero double-spend ledger corruptions.Aegis solves critical FinTech threats: thundering herd surges during payroll days, network latency double-spends, and batch process collisions. Aegis constructs a 4D state vector ($x_t = [\text{Queue Depth}, \text{Thread Occupancy \%}, \text{DB Saturation \%}, \text{Retry Rate}]^T$) and uses Frequent Directions matrix sketching to track subspace shifts via Incremental Singular Value Decomposition (SVD). Projecting telemetry onto a healthy subspace matrix ($P_S = V_k V_k^T$), Aegis calculates residual norm error:$$\text{Residual Norm} = \Vert{} (I - P_S) x_{\text{norm}} \Vert{}_2$$Spikes above 15.00 trigger immediate remediation.The Agentic AI swarm features five specialized NitroStack agents: PRIME (orchestration), ATLAS (traffic engineering), CERBERUS (security), LIBRAN (micro-reconciliation), and HERMES (SOC2 compliance). The swarm executes three defense shields: Single-Flight Query Coalescing to reduce database read locks by 92%, SHA-256 Idempotency Intercepts to block duplicate payment retries, and QoS Admission Control to guarantee 90% bandwidth for critical wire transfers. Built on Node.js, TypeScript, and Next.js 14, Aegis delivers self-healing FinTech infrastructure.
TradeMind AI is a multi-agent trade break reconciliation copilot built for post-trade operations teams. Instead of analysts manually cross-checking trade records, market data, portfolio positions, risk metrics, and compliance rules across six disconnected systems, TradeMind AI orchestrates specialized AI agents — each an expert on one domain (portfolio, market data, risk, trade records, compliance, and team communication) — that work together over the Model Context Protocol (MCP) to spot mismatches, investigate root causes, and surface a clear, actionable resolution, with automatic Slack notifications to the right people. Built on NitroStack, it consolidates 18 tools across 6 domains into a single deployable MCP server.
An autonomous AI-powered Candidate Intelligence Agent built on NitroStack that helps recruiters evaluate candidates by consolidating information from multiple sources into a single verified profile. The system retrieves live candidate data through the GitHub API, Portfolio combines it with resume information, detects inconsistencies across sources, assigns explainable confidence scores to each field, and generates a recruiter-friendly hiring summary with evidence-backed recommendations. Designed using NitroStack's agentic architecture, the solution leverages tools, resources, live API integration, and robust error handling to automate candidate verification, improve hiring decisions, and reduce manual screening effort.
When a new hire joins, HR and IT have to manually create their email, add them to various different channels, assign training modules, and send welcome emails which is a tedious, multi-step process. Conversely, when an employee leaves, they often miss revoking access to some tools, creating abandoned accounts that pose a massive security vulnerability and waste subscription money.
AutoBoard.ai eliminates this issue by operating multi-step orchestration and full lifecycle state management. The AI has to take a single, simple command and autonomously break it down into multiple sequential actions whether building a digital identity from scratch or securely dismantling it. It makes sure that all records of the user are added to their respective accounts and vice versa, where all records are deleted.
Rampo is an AI-powered proactive customer support platform for digital banking, built using MCP and NitroStack. Instead of waiting for customers to raise complaints or contact support, Rampo continuously monitors customer interactions to identify signs of confusion or frustration—such as repeated errors, failed transactions, multiple retries, prolonged inactivity, or rage clicks. When a potential issue is detected, the platform proactively alerts support agents or AI assistants, enabling them to intervene in real time with contextual guidance and navigation assistance within the customer's active banking session. By helping customers complete their tasks before frustration escalates into a support request or complaint, Rampo reduces customer effort, lowers support costs, improves transaction success rates, and delivers a smoother, more personalised banking experience.
AI-powered retail location intelligence. Enter business type, city, budget & radius — RetailMind analyzes competitors, demographics, and footfall potential across candidate zones via an MCP multi-agent architecture, then ranks them with an explainable 0–100 Opportunity Score. Built with TypeScript, NitroStack, Geoapify & WorldPop.
ealthOS AI is an AI-powered personal healthcare assistant designed to simplify and enhance the way individuals manage their health. The platform securely stores and organizes medical records, prescriptions, diagnostic reports, and lab results in one centralized location with strong privacy and data protection. Using advanced artificial intelligence, it analyzes prescriptions and medical reports to identify potential drug interactions, explain complex medical terminology in simple language, and provide personalized health insights based on the user's medical history. HealthOS AI also tracks medications with intelligent reminders, helps users schedule and manage doctor appointments, sends timely notifications for medicine intake, health checkups, vaccinations, and follow-up visits, and offers AI-driven recommendations for maintaining a healthier lifestyle. By integrating secure health record management, smart automation, predictive analytics, and personalized healthcare guidance into a single platform, HealthOS AI empowers users to make informed decisions, improve medication adherence, and proactively monitor their overall well-being while reducing the complexity of day-to-day healthcare management.
NitroGuard is a real-time AI safety gateway for autonomous mobile robots (AMRs) operating in dynamic factory environments, addressing an emerging challenge in physical AI and industrial automation. As large language models are increasingly used for natural-language mission planning in embodied AI systems, a structural gap remains: LLMs lack deterministic spatial awareness and cannot inherently guarantee collision-free execution. Existing industrial safety standards, such as ISO 10218:2025, were built for conventional control architectures and have not yet been extended to account for AI-generated motion intent.
NitroGuard addresses this by acting as an MCP-native interception layer between AI-driven planning and physical actuation. Every movement command proposed by the LLM is evaluated by a Control Barrier Function (CBF) safety engine before execution, mathematically constraining the trajectory outside defined hazard boundaries in real time. The language model never holds direct write access to the robot. Mission planning is grounded in live MCP Resources — factory layout, hazard map, robot state — before a target is proposed. The raw versus safety-corrected trajectory is rendered live in 2D/3D, then dispatched to a MuJoCo physics simulation for execution.
Built on the full NitroStack framework: Tools for mission execution and emergency stop, Resources for environmental grounding, a Prompt template for safe-navigation workflows, Guards and rate-limiting on execution, and an interactive trajectory-viewer Widget.
Intended for robotics teams integrating LLMs into control loops, industrial manufacturers adding natural-language interfaces, and integrators needing auditable safety assurances as AI-robot regulation develops.
EnterpriseAI is an agentic Natural-Language-to-SQL platform that lets anyone in an organization query a complex ERP database — spanning 500+ tables — just by asking a question in plain English, no SQL knowledge required.
Under the hood, a 10-step agentic pipeline normalizes the query, extracts intent, identifies the right domain and tables from the schema, plans a safe read-only query, generates SQL via an LLM (Groq/LLaMA), validates it against anti-hallucination and safety filters, self-heals on errors, executes it, and returns a natural-language summary with insights.
We packaged the entire pipeline as an MCP (Model Context Protocol) server, exposed over HTTP — so it's not locked into one app. Any MCP-compatible AI client (Claude, agents, internal tools) can plug in and query enterprise data conversationally, turning a rigid ERP system into something employees can actually talk to — cutting the time from "I need this report" to an answer from minutes of manual SQL/BI work down to seconds
What it does: A student speaks a short reflection on what they just learned. Their speech is transcribed into text, then one agent identifies any misconceptions using a fixed taxonomy. A second agent searches the actual course material for the correct explanation and verifies it with the relevant page before sending it back. If it cannot find a grounded citation, it does not generate a correction. Instead, the case is escalated to the faculty. Across an entire class, deterministic code groups similar misconceptions into clusters, and another agent traces each cluster through a prerequisite graph to identify the underlying concept students are struggling with. A week later, the system checks the same misconception again and reports whether it has been resolved.
Who it's for: Primarily faculty, who often discover what students misunderstood only weeks later through exam scores that reveal little about the reason behind the mistakes. It also benefits students by creating a learning roadmap based on demonstrated mastery rather than self-reported interests.
What makes it special: Speaking is more natural and requires less effort than writing. Students are often more honest and expressive when they explain what confused them out loud, making speech transcription an essential part of capturing meaningful learning signals. Each agent runs behind its own MCP server with access only to the tools required for its specific role, so the separation between a student's raw voice reflections and the faculty's aggregated insights is enforced through system permissions rather than policy alone. Every correction must be backed by evidence from the course material before it reaches the student, preventing unsupported responses. Finally, the system closes the feedback loop by checking a week later whether the misconception has actually been corrected instead of assuming the intervention was successful.
ModelSmithAI is an autonomous, multi-agent system that turns a plain-English request into a trained, security-scanned machine-learning model in minutes. Six coordinated agents gather open-licensed images from the web, verify them, train a classifier, diagnose their own failures, gather better data, and deliver a model in the safe SafeTensors format with a PDF audit report. Built on NitroStack MCP, the same agents run behind a web app and an orchestratable AI chatbot.
Who it's for: teams and individuals with no ML expertise, engineers facing niche tasks with no ready-made dataset, and MLOps/security teams who need to vet model files before running them.
What's special: it's genuinely self-improving — the Diagnostician reasons about which classes fail and what data fixes them — and it treats security as a guarantee, converting every model to a format that physically cannot execute malicious code.
The advantage: what normally takes an expert days or weeks, ModelSmithAI does in minutes — with no dataset, no labeling, and no code. Unlike AutoML tools that assume you already have data, it sources the data itself, improves through its own reasoning, and guarantees the final model is safe to use.
Give an agent a CRM, a messaging system, and a billing API, and eventually it will do something it can't take back - mid-plan, after five other steps have already run. This project wraps every tool call an agent makes inside a journaled, LIFO-reversible transaction, classifies each action by how reversible it actually is, and gives you an honest report - including the parts that can't be undone - when something needs to be rolled back.
Built as an MCP server on NitroStack, with a live timeline widget and a 7/7-passing acceptance suite.
When an Indian earner dies, the money doesn't disappear — the family just can't find out how to claim it. Each institution knows about its own asset and explains nothing about the others, so families discover the process one rejection at a time, and smaller assets get abandoned entirely. Claim Navigator is a machine-readable rulebook that any AI assistant can consult. The family describes what they know; deterministic tools return the applicable claim route under the RBI Settlement of Claims Directions 2025 and Indian succession law, the consolidated document checklist, the expected cost, and the realistic timeline — each output labelled as regulatory, institution policy, or estimate.
TrustLayer AI MCP Server — Edge-ready Model Context Protocol (MCP) risk verification engine built on NitroStack. Exposes 9 tools and 3 resources for real-time claim extraction, contradiction detection, and probabilistic risk evaluation over HTTP/SSE.
Patch Made In Heaven acts as a real-time "Stack Overflow" built exclusively for AI agents.
When an AI developer agent encounters a runtime error, stack trace, or environment failure, it queries Patch Made In Heaven natively via the Model
Context Protocol (MCP).
• Instant Retrieval: If a matching fix exists, the platform uses hybrid BM25 + dense vector semantic search to return an environment-matched,
execution-verified patch diff in under 1 second.
• Automated Sandbox Verification: If no fix exists, the agent resolves the issue, validates the patch inside an isolated Docker sandbox against real
STDOUT/STDERR logs, and publishes the verified solution back to the global index.
• Zero Token Waste: Once any agent verifies a bug fix, every other agent on the network instantly inherits the solution—eliminating repetitive LLM
reasoning, token waste, and redundant debugging cycles.
**MigrateEase** is an AI-powered relocation and visa-processing platform that carries the heavy, rules-heavy work of moving across borders from end to end — figuring out the exact requirements for each client's specific destination and visa type, guiding them through a conversational, validated onboarding, collecting and checking every document automatically, recommending the right property brokers, and tracking the whole journey, all while keeping a human expert in control of every consequential decision. It's built for two audiences at once: people relocating abroad who face the confusion of "what do I actually need to submit?" for a move whose rules change from country to country and visa type to visa type, and the relocation companies serving them, whose operations teams run everything through WhatsApp, spreadsheets, and tribal knowledge and need to scale without linearly adding headcount or risking costly, preventable errors. What makes it special is that instead of treating visa paperwork as a fixed checklist, MigrateEase understands the real problem — requirements are a moving target defined by destination × visa type × applicant profile, where one wrong detail can derail an entire move — so it builds a dynamic, country-aware knowledge base on demand, runs on an agentic human-in-the-loop model where automation handles the repetitive 90% while an expert approves the risky 10% with a full audit trail behind every decision, and meets people where they are with multilingual, multimodal onboarding, feeling less like filling out forms and more like having an expert guide who never drops the ball.
TrialSync AI+ is an end-to-end multi-agent AI copilot designed to streamline clinical trial matching, eligibility screening, and patient retention.
What it does:
• Automated Trial Discovery: Connects live clinical trial protocols (e.g., ClinicalTrials.gov) with structured patient profiles.
• Precision Eligibility Screening: Leverages multi-agent pipelines to evaluate unstructured clinical notes and biomarkers against complex inclusion/exclusion criteria.
• Conversational Copilot: Equips clinicians with a Doctor AI Copilot to query trial protocols, laboratory thresholds, and treatment recommendations in real time.
• Proactive Patient Safety & Monitoring: Tracks patient adherence, logs adverse events, and monitors toxicity levels to alert care teams before side effects escalate.
• Executive Analytics: Generates automated clinical summary reports and sponsor-level analytics for enrollment funnels and retention stats across sites.
Who it is for:
Designed for clinical researchers, oncologists, trial coordinators, trial sponsors, and trial participants.
What makes it special:
TrialSync AI+ bridges the gap between complex research protocols and clinical workflows through specialized AI agents working in tandem. It introduces a dual, synchronized ecosystem—a physician workstation for protocol decision-making alongside a dedicated patient portal for transparent care tracking—all protected by role-based access control.
During medical emergencies, choosing the nearest hospital is not always the safest option. Current navigation systems consider only distance and traffic, ignoring critical factors such as hospital specialization, real-time ICU/ER bed availability, and emergency treatment capabilities. As a result, patients may be taken to hospitals that cannot provide the required care or have no available capacity, leading to delays, secondary transfers, and increased risk during the critical Golden Hour.
AuraHealth addresses this problem by using an Agentic AI system built with NitroStack MCP to analyze the patient's condition, identify the most suitable hospital based on medical capabilities and live resource availability, optimize the fastest route, and reserve an emergency bed before the patient arrives. This enables faster, smarter, and more reliable emergency care decisions.
**Converra One – Project Description**
**Converra One** is an AI-powered enterprise productivity platform designed to simplify workplace communication, collaboration, and task management by integrating multiple business applications into a single intelligent workspace. Modern organizations rely on tools such as Gmail, Google Calendar, Slack, GitHub, Notion, and Discord for daily operations, which often results in information overload and frequent context switching. Converra One addresses this challenge by providing a unified dashboard that centralizes all essential workplace information.
Built using **NitroStack Studio**, **React**, **Next.js**, **TypeScript**, and the **Model Context Protocol (MCP)**, the platform enables seamless interaction between AI agents, backend services, and frontend widgets. It features a **Unified Inbox** that consolidates messages from multiple platforms, an **AI Smart Search** for retrieving information using natural language, an **AI Daily Briefing** that summarizes emails, meetings, and tasks, and an **AI Smart Reply Composer** that generates context-aware email responses.
The platform also integrates calendar management and task tracking, allowing users to manage schedules, monitor deadlines, and prioritize work efficiently. Through MCP-based architecture, Converra One supports scalable integration with enterprise tools while maintaining a modular and extensible design.
By leveraging **Agentic AI**, the platform intelligently prioritizes communications, automates repetitive workflows, and provides actionable insights to improve productivity. Converra One reduces information overload, minimizes application switching, and enhances decision-making, making it a comprehensive AI-powered digital workspace for modern enterprises.
Golden Hour is an AI-powered, MCP-enabled cyber fraud response platform that transforms fraud complaints into investigation-ready case packets through a deterministic three-agent workflow. It automatically classifies fraud, assigns cases to the appropriate authorities, and provides jurisdiction-specific legal guidance with explainable recommendations. By combining specialized AI agents, Model Context Protocol (MCP), and human-in-the-loop decision making, Golden Hour enables faster, transparent, and more reliable cyber fraud investigations while ensuring all final enforcement decisions remain with authorized personnel.
This project is an AI-powered financial resilience assistant that helps individuals, freelancers, gig workers, and small business owners make smarter financial decisions before problems become crises. It continuously identifies unnecessary expenses, understands income stability even when earnings are irregular, and creates personalized action plans for unexpected financial emergencies. Instead of simply displaying past transactions, it provides proactive insights and practical recommendations to improve financial security. Its unique focus on resilience, rather than just budgeting or expense tracking, makes it a forward-looking financial decision-support system.
DeploySense AI is an agentic AI reasoning engine built on top of the NitroStack MCP Framework. It acts as an automated Incident Commander that investigates, diagnoses, and documents post-deployment production failures using 8 integrated MCP tools and state-driven simulation scenarios.
Today an agent that needs to run code gets a shell, a container, or a cloud key. The grant is all-or-nothing and invisible: agents ask for 16 GB and root "to be safe" because the alternative is failure, and blocked agents retry against an opaque wall with no machine-readable reason to reason about.
BouldersGate makes the grant itself a conversation.
request_compute evaluates a request against human-authored policy and returns one of three answers — and creates no environment in the process, so an agent can find the edge of what it may have before committing to anything. Quantitative excess comes back as reduced (16384 MB → 4096 MB), scope excess as narrowed (unrestricted egress → none), each as structured data with path, requested, granted, and reason, so the caller can replan against the real grant. Capabilities with no safe reduced form — privileged, hostFilesystem, dockerSocket — are denied outright with no offer attached, never quietly downgraded to something adjacent. Policy is never exposed as an MCP resource: agents learn the shape of their grant through outcomes, not by reading the ruleset.
Multi-tenancy is enforced, not assuma distinct opaque identity — agent_plus a SHA-256 fingerprint — so the responses. Environments belong to that identity, and ownership is checked o_environments, andrelease_environment: one agent canno another's environment, and the testsuite asserts exactly that. list_audlling agent's own decisions, withcommand contents and credentials nevGrants are enforced rather than prom image digest resolved from the OCIregistry at negotiation time, and acexact digest — a mutable tag cannot be
re-pointed between the promise and ts with zero inherited environment
variables, egress removed, and the fironment's own workspace, on a
container backend where a Docker daeess backend where none does. Both are
verified by assertions in the test sve NitroCloud deployment.
Our system provides an end-to-end AI-powered healthcare workflow that begins by accepting a patient's laboratory report in the form of text, image, or PDF. Image and PDF reports are first processed through OCR or PDF parsing to extract raw text, which is then standardized into structured laboratory parameters by the parse_labs module. The extracted values are analyzed by flag_critical to classify each test as Normal, Borderline, or Critical using reference ranges. Based on the analysis, suggest_doctor recommends the most appropriate specialist while considering whether the patient should continue with an existing doctor or consult a new one. The patient can then view available appointment slots, book a consultation, and receive a confirmation with intake details. Finally, share_summary generates a concise, patient-friendly summary that can be shared via messaging platforms. For a seamless experience, run_full_triage automates the complete workflow in a single call while maintaining patient history and trend analysis across multiple visits, and secure authentication ensures role-based access for both patients and doctors.
ResearchRadar MCP is an open-source Model Context Protocol (MCP) server that enables AI assistants to discover, analyze, and organize academic research through natural language. It helps developers, students, researchers, and knowledge workers search for relevant papers, summarize findings, compare research, and build curated reading lists directly within MCP-compatible clients such as Claude, Cursor, and other AI tools. Built with TypeScript and designed for extensibility, ResearchRadar simplifies literature exploration by turning scattered academic sources into structured, actionable insights through a standard MCP interface. It also helps identify the research gap and facilitates a comprehensive literature review. It helps innovators and entrepreneurs to find new ideas that could be commercialized.
ClaimShield is a data-driven orchestration engine that guides policyholders through complex insurance claims while actively preventing AI hallucinations. It serves everyday consumers who need legally accurate guidance to avoid unfair claim denials, as well as insurance companies looking to safely automate claim triage and detect price gouging. What makes ClaimShield truly special is its Model Context Protocol (MCP) multi-agent architecture. Instead of letting a generative AI guess or fabricate legal advice, the system strictly forces the AI to fetch verified regulatory clauses and hardcoded market repair costs from a secure local database before it generates any action plans or appeal letters.
What it does: Decision Twin is an agentic AI system for manufacturing plants that automates operational decisions when a machine reports an anomaly, abnormal vibration, temperature, pressure, etc. Instead of one model producing a single verdict, it launches a full multi-agent negotiation: Sensor, Maintenance, Memory, Production, Quality, Inventory, and Finance agents gather evidence in parallel, then a Devil's Advocate and Safety agent stress test the leading proposal against risk thresholds and hard SOP rules (like budget caps). If a proposal gets challenged, the system renegotiates and re run a quantitative scenario simulation comparing options like immediate repair, delayed repair, or reduced capacity operation, each scored on cost and risk. It converges on a final recommendation with a confidence score, an auto generated work order, and a supervisor notification flagging whether approval is required.
Who it's for: Plant managers, maintenance engineers, and operations teams who need fast, defensible decisions on machine downtime and repair timing without manually cross-checking sensor data, incident history, spare parts inventory, and financial impact by hand every time an alert fires.
What makes it special: It's not a black-box classifier but it's an auditable negotiation. Agents can genuinely veto and challenge each other (example: Safety blocking an "immediate repair" that breaches a budget cap, forcing a pivot to "delay repair"), and every round of that back and forth is logged. It transparently reconciles conflicting data sources (user-submitted readings vs. live sensor telemetry), showing exactly which value was trusted and why. Every decision comes with a full trace agent-by-agent reasoning, precedent incidents consulted, vetoes, and simulation scores, giving operators a decision they can actually inspect and trust, not just accept.
Finly is a multi-agent personal finance assistant for two groups: people who are sold financial products they were never taught to evaluate, and people whose income isn't a salary.
It explains the jargon used when money is being sold to you, shows what stopping a policy early would really cost, scores how much a seller actually disclosed, builds ordered plans for a first job or a house, and checks affordability against your weakest month rather than your average.
No agent ever calculates — every figure comes from tested tools. It refuses to name products to buy, because that's regulated in India.
SORCME is a MCP Server that connects multiple tools related to cryptocurrency together by fetching current crypto prices from various sellers and runs the SOR algorithm to route the purchases to provide maximum returns for cost.
MedLens is an agentic MCP server hosted on NitroStack Cloud that queries public health databases like openFDA (mcpmarket.com) and RxNorm to provide source-cited, conversational medicine insights and act as a live medication schedule companion.
Campus Sārathi is an agentic academic assistant that unifies a student's calendar, email, tasks, and campus updates into one intelligent MCP-powered interface — turning scattered information into a single prioritized daily plan.
CloudOps OS is an agentic AI framework built on NitroStack MCP that orchestrates four specialized LLM agents (Resource Routing, Cost Optimization, Policy Governance, Platform Release) to automatically review and remediate Terraform infrastructure code before deployment.
Modern LLMS are prone to manipulative exploits like prompt poisoning and jailbreaks. finding and fixing these vulnerabilities are time consuming and human resource intensive. We intentend to optimize this search by automating the pentesting through MCP servers by using AI agents.
Every day, millions of people send money across borders—to support their families, pay university tuition, cover medical emergencies, or manage business payments. But choosing the right remittance provider isn't as simple as it should be. Users are often forced to compare multiple platforms, understand fluctuating exchange rates, calculate hidden fees, estimate delivery times, and navigate compliance requirements, all while hoping they're making the right financial decision.
That's where RemitWise AI comes in. We've built an intelligent multi-agent platform that takes the complexity out of international remittances. Instead of making users do all the research themselves, our AI agents work together to analyze the transfer, compare providers in real time, evaluate compliance requirements, and recommend the best option based on the user's priorities—whether that's the lowest cost, fastest delivery, or best exchange rate.
But we didn't stop at recommendations. We wanted users to trust every decision our platform makes. That's why RemitWise AI provides transparent AI reasoning, live provider comparisons, interactive visualizations, a real-time agent execution pipeline, compliance insights, and exchange-rate predictions that help users decide not just where to send money, but also when to send it.
At its core, RemitWise AI isn't just another remittance tool—it's an intelligent financial decision-making platform that empowers users to transfer money with confidence, transparency, and peace of mind.
FinVerse is an AI-powered financial orchestration platform designed to empower India's gig workers and micro-merchants with secure, inclusive, and intelligent financial services. By leveraging the Account Aggregator ecosystem, machine learning, cryptographic fraud detection, and adaptive micro-repayment mechanisms, FinVerse enables fair credit assessment based on real cash flows instead of traditional credit scores. The platform also safeguards families through an AI-driven succession engine that simplifies asset discovery and claim processing, ensuring financial security across generations.
AISRA is an AI-powered operations assistant for home bakeries built using NitroStack's Model Context Protocol.
It demonstrates how MCP tools can be orchestrated to automate inventory analysis, supplier recommendation, decision support, and RFQ generation through a conversational workflow.
Sentinel is an AI-powered cybersecurity assistant built using the NitroStack MCP Framework. It helps organisations handle security alerts automatically, reducing the time security teams spend on repetitive investigations.
Whenever a suspicious activity is detected, Sentinel analyses the alert, identifies how serious it is, explains what happened in simple terms, and suggests the next actions. Instead of manually checking hundreds of alerts, security analysts receive a clear, structured investigation report within seconds.
To make investigations easier, Sentinel also includes an interactive dashboard that displays the threat level, summary, and recommended actions in an easy-to-understand format. It is further integrated with NitroChat and the ChatGPT MCP Plugin, allowing users to investigate incidents using natural language conversations
Sentinel transforms complex cybersecurity alerts into simple, actionable insights, helping security teams respond faster while reducing manual effort and improving efficiency.
Key Highlights
AI-Powered Tier-1 SOC Assistant
Automated Security Alert Triage
Threat Severity & Risk Assessment
MITRE ATT&CK Mapping
Threat Intelligence (IP & File Hash Analysis)
Automated Incident Report Generation
Interactive SOC Dashboard
NitroChat & ChatGPT MCP Plugin Integration
Faster Incident Response with Reduced Manual Effort.
IGNITE AI is an AI-powered startup intelligence platform that transforms raw business ideas into investor-ready venture insights. By leveraging sequential AI orchestration, the platform analyzes startup concepts, evaluates market opportunities, identifies competitors, assesses risks, and generates strategic recommendations—all within a single interactive workflow.
Designed for founders, innovators, students, and entrepreneurs, IGNITE AI accelerates the journey from idea validation to venture planning by providing data-driven business intelligence, opportunity scoring, market research, and actionable next steps.
A NitroStack MCP-based Agentic Underwriting Framework integrating enterprise intelligence with deterministic policy evaluation for explainable, auditable, and counterfactual loan decisions.
Aide is a Model Context Protocol (MCP) server built with NitroStack. It serves as an intelligent backend router and tool provider for AI agents (like ChatGPT, Claude Desktop or Cursor), exposing enterprise capabilities such as task delegation, calendar scheduling, admin policy checking, and communication routing with Slack and Discord integration.
MediGuardian AI is a secure multi-agent healthcare platform that helps doctors and caregivers make safer treatment decisions. Instead of simply storing medical records, it combines a patient's medical history, current medications, laboratory reports, genetic profile, and allergies to identify potential risks before treatment.
The platform performs Drug–Drug, Drug–Disease, Drug–Gene (Pharmacogenomic), and Laboratory-based interaction analysis to detect unsafe prescriptions and recommend safer alternatives. For example, it identifies genetic variants such as CYP2C19, which can prevent medicines like Clopidogrel from becoming effective, enabling personalized medication recommendations. It also detects abnormal dosages, medication conflicts, and verifies medicine authenticity using official drug registries to help identify counterfeit or mislabeled drugs.
MediGuardian AI maintains a Health Memory that continuously organizes patient information and generates a dynamic Health Score to provide a quick overview of overall health and clinical risk.
Built with a Zero Trust Security Architecture, every interaction between AI agents is authenticated, authorized, encrypted, and audited, ensuring patient privacy while enabling intelligent clinical decision support.
A production-grade context engine that connects an AI agent directly to fragmented enterprise data silos (Jira, Slack, Notion, HR policies, and technical architectures). It uses semantic vector search to allow employees to query internal documentation, check project ownership, or summarize meeting action items using natural language. Designed for enterprise developers and internal staff. What makes it special is its strict adherence to role-based access control (RBAC) and its ability to autonomously execute multi-step workflows, like diagnosing an incident and automatically generating a high-priority IT ticket based on the findings.
ApexTune is an AI-powered personal PC optimizer built as an MCP server. It lets an AI agent inspect your system in real time — running processes, startup apps, disk space, and network status — and safely fix what's slowing you down: killing runaway processes, disabling unnecessary startup apps, clearing temp files, and more.
It's for everyday PC users who want the benefits of tools like Task Manager and CCleaner, but explained in plain language and driven by conversation instead of digging through menus — just describe the problem ("it's slow", "something feels off") and the agent diagnoses it for you.
TokenSlash is an advanced solution for prompt optimization and model selection that uses Artificial Intelligence as its driving force, leading to considerably lower costs, reduced tokens consumption, and shorter time spans for Large Language Model software. It works as an intelligent layer between the user and the AI software, analyzing the prompts before it is deployed.
The system calculates tokens consumed, determines the type and complexity of the task, eliminates redundant or unnecessary content, and reformulates the prompt into a comprehensible form. Thereafter TokenSlash compares a number of AI models against their anticipated output quality, price, time for execution, and retry probability, in order to provide the user with the best possible solution in terms of cost and effective performance.
TokenSlash is advantageous for developers, startups, teams working with products, departments engaged in customer support, enterprises, and teams responsible for governance. It is helpful for chatbots, coding assistants, document processing systems, support automation tools, and other systems dealing with AI technologies on a daily basis.
TokenSlash is created based on NitroStack SDK and Model Context Protocol and works with NitroCloud, NitroChat, Cursor, and Claude Desktop. So AI work becomes easier and more transparent.
Trade Matcher reconciles trades between two mock trading systems (System A and System B) — detecting matches, price/quantity discrepancies, and unmatched trades — a common and tedious problem in financial back-office operations. Beyond simple matching, it uses an LLM to investigate why a mismatch exists and propose corrections for anything it can't confidently explain, while keeping a human explicitly in the loop for every final decision.
load_trades — loads mock trades from System A, System B, or both
match_trades — pairs trades by symbol, flags anything outside price/quantity tolerance, or missing entirely from one side
investigate_break — an LLM call reasons about why the mismatch might exist (FX timing, settlement windows, or genuinely unexplained)
resolve_or_escalate — explained breaks are auto-resolved; unexplained breaks are escalated for human review
propose_correction — for escalated breaks only, a second LLM call conservatively proposes a specific fix (which field, which system, what value) with reasoning — but explicitly never applies it
run_reconciliation — orchestrates the full pipeline end-to-end in a single tool call and renders results in a live dashboard widget
HUNT3R-T is an autonomous threat hunting agent that eliminates the 197-day breach detection gap.
Unlike current tools that alert humans and wait, HUNT3R-T hunts actively, simulates attacker paths in a digital twin, and pre-emptively neutralizes threats — all in under 4 minutes with zero human intervention.
Key differentiators:
• Autonomous hunting: No alert required. The agent searches for threats across SIEM, EDR, and threat intel simultaneously.
• Digital twin simulation: Every response is battle-tested in a parallel universe before touching production.
• Explainable autonomy: Full causal chain for every action — no black boxes, no compliance risk.
• Temporal reconstruction: Reconstructs attack timeline to find patient zero and predict next moves.
Result: 197 days of attacker dwell time → 4 minutes. Not faster response. Prevention.
Built with NitroStack MCP architecture for hackathon deployment.
ArogyaAI OS is an AI-powered Hospital Intelligence Platform built with NitroStack and the Model Context Protocol (MCP) that unifies fragmented hospital services into a single conversational AI experience. Instead of requiring separate systems for appointments, doctor discovery, laboratories, pharmacies, bed management, emergency coordination, and hospital analytics, ArogyaAI OS enables patients and healthcare staff to access these services through one intelligent assistant.
The platform combines 15 MCP tools, interactive Generative UI widgets, live MongoDB data, and specialized AI workflows to support symptom triage, doctor recommendations, appointment management, emergency response, pharmacy and laboratory services, hospital capacity monitoring, and executive decision-making. Its modular architecture allows AI to safely orchestrate hospital operations while keeping business logic within structured MCP tools rather than relying on free-form responses.
Designed for the NitroStack AI Buildathon 2026, ArogyaAI OS demonstrates how agentic AI can improve healthcare accessibility, streamline hospital workflows, and provide operational intelligence through a scalable, MCP-native architecture.
ForgeOps is an explainable, multi-agent manufacturing decision-intelligence platform that turns fragmented factory data into evidence-backed corrective actions. It addresses a major industrial challenge: traditional monitoring systems can detect faults but often cannot explain their causes, compare possible interventions, or justify recommendations.
A Python orchestrator coordinates four specialized AI agents. The Planner defines the investigation, Research retrieves operational evidence, Analysis identifies causal relationships and evaluates scenarios, and Execution generates the final recommendation and business impact.
The agents use NitroCloud’s deployed NitroChat inference layer with a configured Gemini model. They interact with 16 MCP operations—11 direct tools and five task workflows—to access MES events, sensor telemetry, maintenance history, quality inspections, supplier constraints, production timelines, causal graphs, simulations, and business-impact calculations.
ForgeOps reconstructs incidents, connects anomalies across production stages, separates symptoms from root causes, and compares actions using predicted yield, cost, effort, downtime, and confidence. Engineers can investigate incidents through natural-language questions and what-if simulations.
The React frontend communicates with the Python orchestration backend through HTTP APIs. The backend combines LLM reasoning with controlled MCP data retrieval and returns structured results containing evidence references, assumptions, confidence scores, agent status, and a live tool trace. Human approval remains the final gate before action.
ForgeOps is not simply an LLM dashboard; it is an auditable agentic decision system that helps manufacturers reduce downtime, prevent costly misdiagnosis, improve production yield, and make faster, explainable operational decisions.
SemiComply is a Unified Semiconductor Lifecycle MCP Server built with NitroStack. It connects five fragmented semiconductor enterprise domains — Cadence Design, MES Telemetry, STDF Quality/Yield, Product Specs, and SAP Shipping & Trade Compliance — into an agentic intelligence hub with live interactive widgets.
WHO IT IS FOR:
Semiconductor engineers, quality managers, fab operators, and trade compliance teams who traditionally waste days manually querying 4–5 siloed enterprise databases to answer cross-domain questions like "Why did this wafer lot's yield drop, and is the shipment blocked at customs?"
WHAT IT DOES & WHAT MAKES IT SPECIAL:
1. Autonomous Cross-Domain Orchestration: Unlike basic API aggregators, our orchestrator (`analyze_yield_root_cause`) correlates sensor excursions (e.g. etch pressure spikes) with STDF bin failures (e.g. BIN_12_LEAKAGE), rules out design flaws, and evaluates downstream customs compliance in a single request.
2. Full Wafer Genealogy Timeline: `trace_wafer_genealogy` constructs structured, phase-ordered wafer timelines (Product → Design → Fab → Test) and physical global trade routes.
3. Automated Stakeholder Incident Escalation: `notify_incident_stakeholders` classifies severity, resolves roles from live sheets, sends branded HTML email reports via SMTP, and records an audit log.
4. Live Google Sheets Backend: Zero-database friction. Data is fetched live from published Google Sheets with a 30s cache, making it instantly editable by non-technical teams without code deployment.
5. Interactive NitroStack Widgets: Embedded Next.js visual widgets (`lot-yield-timeline` & `wafer-lifecycle`) render rich charts and interactive supply chain maps directly inside NitroStudio.
Care Mediator is a neutral AI layer that sits between hospitals, insurers,
and patients during a medical insurance claim — the exact moment financial
disputes happen. Instead of the patient being stuck guessing whether a
hospital overbilled or an insurer wrongly denied a claim, our system runs
automated agents that cross-check every submission against government CGHS
rate benchmarks and the insurer's own claim record, flagging inconsistencies
before either party can act on them.
When a coverage gap exists, our Lender agent doesn't just list financing
options — it computes true effective annual rate (not headline APR) to
automatically flag predatory loans dressed up as "low interest."
Built as five specialized MCP agents (Hospital, Insurer, Lender, Objectivity,
Orchestrator) sharing one case record, so hospital, insurer, and patient
always see identical, tamper-evident data — no version conflicts, no
selective disclosure.
Who it's for: patients navigating opaque medical billing, insurers wanting
pre-vetted claims, and hospitals needing audit-proof submissions. What
makes it special: it's not a chatbot bolted onto insurance — it's a
transparent reconciliation engine where every number is independently
verifiable against a neutral benchmark before a human ever has to trust
either side's word.
MCP zero-trust gateway that sits in front of every internal MCP server, detects tool-poisoning/metadata drift in real time, and maintains a cryptographic provenance ledger of every agent tool call
Lannex is an AI money assistant that tracks your spending, splits bills from receipt photos, and chases friends who owe you through natural conversation.
It also spots idle cash from your daily transactions and nudges you to invest it, while giving you a live view of your balance, debts, and portfolio in one dashboard.
Resilio City is an AI-powered urban resilience platform that helps authorities assess, simulate, and recover from infrastructure disasters. It loads a city's road network, simulates disaster scenarios such as floods, earthquakes, and landslides, and analyzes their impact on critical infrastructure. Using specialized MCP agents, it automatically generates prioritized recovery plans with budgets, manpower estimates, repair phases, emergency routes, and actionable recommendations. The platform enables faster, data-driven disaster preparedness and recovery planning for smarter and more resilient cities.
An MCP that helps manage, organize, and write content in your Obsidian vault. It powers an automated AI research and document conversion pipeline that transforms YouTube videos, PDFs, and code into structured study notes, performs live web research, generates practice quizzes, and builds interconnected Obsidian Knowledge Graphs.
RIVA (Ride Intelligence & Virtual Assistance) is an AI-powered mobility copilot designed to make urban transportation smarter, safer, and more personalized. Unlike traditional ride platforms that only display fares, RIVA understands user intent, analyzes travel requirements, and provides intelligent recommendations through natural language conversations.
Built using MCP (Model Context Protocol), RIVA connects multiple mobility services, mapping intelligence, and AI-driven decision-making into a unified assistant. Users can simply describe their journey, and RIVA can compare ride options, optimize routes based on travel time and traffic conditions, estimate fares, and recommend the best choice based on price, availability, preferences, and convenience.
RIVA introduces an AI Travel Concierge experience where users can plan complete journeys instead of only booking rides. For example, a user travelling to an airport can ask RIVA to plan their trip, and the system can determine the ideal departure time, analyze traffic conditions, suggest suitable rides, estimate future fares, and provide a complete travel plan.
The platform enhances mobility through features such as map-based ride visualization, traffic-aware route optimization, multi-stop trip planning, simulated carpooling, future fare prediction, safety assistance with emergency location sharing, and simulated booking workflows.
RIVA also provides a Driver Intelligence mode that helps drivers make smarter decisions by analyzing ride profitability, estimating earnings, identifying high-demand areas, and recommending better ride opportunities.
By combining conversational AI, MCP-based orchestration, predictive analytics, and mobility intelligence, RIVA creates a next-generation transportation ecosystem that improves convenience for passengers and efficiency for drivers.
Why should staying safe from a digital scam require technical knowledge? Every day, millions of people receive messages saying “Your bank account has been blocked,” “Your KYC has expired,” “Click this link immediately,” or “Share your OTP to verify your account.” A single wrong click can expose personal information, banking credentials, or hard-earned money. Yet many existing security solutions expect the user to recognize the threat first, install an application, create an account, visit a website, copy a suspicious URL, or understand technical warnings. For elderly citizens and people with limited digital literacy, the protection itself can become complicated.
ScamShield: India’s AI-Powered Jan Suraksha Bot changes this by asking the user to do only one thing they already know—forward a message. If a suspicious message, URL, bank alert, OTP request, or potential phishing attempt arrives through WhatsApp or SMS, the user simply forwards it to ScamShield, exactly like forwarding a message to a friend or family member.
Behind this simple interaction, Twilio connects the user to a NitroStack-based backend, where MCP-powered AI analyzes the forwarded content for phishing URLs, impersonation, urgency tactics, credential theft, OTP fraud, and social-engineering patterns. Within the same familiar messaging environment, ScamShield returns a clear risk level, explains why the message is suspicious, and tells the user what action to take next.
The core innovation is its zero-friction, forwarding-first cybersecurity model: no separate app, no login, no browser, no extension, and no technical expertise required. Even onboarding follows the same philosophy. The user simply scans a QR code, which automatically opens the correct Twilio WhatsApp conversation with the required joining message already pre-typed—they only tap Send.
Scan Once → Forward Anything Suspicious → Know the Risk → Stay Safe.
AetherOS is an Agentic AI-powered healthcare platform built with NitroStack and Model Context Protocol (MCP), designed to protect and assist patients before, during, and after hospitalization. It orchestrates specialized AI agents and MCP tools for healthcare navigation, billing and pharmacy audits, cashless scheme support, emergency workflows, doctor discovery, insurance intelligence, medical record summarization, and post-discharge recovery—all through a unified patient-centric platform.
SentinelPay AI is an intelligent payment investigation assistant built to help financial teams detect fraud, reduce risk, and make faster, more informed payment decisions. Instead of manually reviewing multiple systems and documents, users can interact with the platform using natural language to retrieve transaction details, verify vendors, analyze invoices, review payment history, evaluate policy compliance, assess financial risk, and generate approval recommendations. By combining structured financial data with AI-powered reasoning through the Model Context Protocol (MCP), SentinelPay AI simplifies complex investigation workflows, improves transparency, and helps organizations make consistent, explainable decisions while saving time and reducing operational overhead.
CampusPilot AI is a multi-agent AI assistant designed to simplify student life by bringing all academic activities into one intelligent platform. It helps students manage class schedules, assignments, attendance, notes, study plans, placement preparation, and deadlines through a natural language conversation. Using MCP tools, resources, and AI agents, CampusPilot can retrieve academic information, summarize study materials, generate quizzes, create personalized study plans, and send smart reminders based on each student's needs. Unlike a traditional chatbot, CampusPilot acts as an autonomous academic assistant that understands context, coordinates multiple tasks, and helps students stay organized, productive, and better prepared throughout their academic journey.
Healthcare today is fragmented. Every hospital stores patient records separately, forcing patients to carry prescriptions, X-rays, scan reports, and laboratory reports wherever they go. If these records are unavailable, doctors may not have complete information about the patient's medical history, leading to repeated tests, delays in treatment, and inconvenience.
To solve this problem, we developed CareVault, an AI-powered Universal Digital Healthcare Ecosystem that securely connects patients and hospitals through one platform.
Let's begin with the Patient Portal.
The patient logs in using their mobile number or Gmail. Once logged in, they can access all their medical records, prescriptions, previous hospital visits, and appointments from one place.
Patients can also book doctor consultations and diagnostic services such as blood tests, X-rays, MRI scans, CT scans, ECGs, and ultrasound scans. Before confirming an appointment, they can view doctor availability, available time slots, the live waiting queue, estimated waiting time, and make payments directly through the application.
To make healthcare even easier, CareVault includes an AI assistant called Health Inspector.
The patient simply says,
'Book an appointment with Dr. Kumar tomorrow morning.'
The AI is not simply responding to the user—it uses our MCP server to communicate with hospital systems, retrieve live data, and perform real actions such as booking appointments, checking doctor availability, retrieving medical reports, and providing live waiting times.
Once the appointment is confirmed, the patient immediately receives the booking confirmation.
Now let's move to the Hospital Portal.
Only authorized doctors, laboratories, radiology departments, and hospital administrators can log in securely. They can upload prescriptions, laboratory reports, X-rays, MRI scans, CT scans, discharge summaries, update doctor schedules, and manage waiting queues.
Problem: Manufacturers run one of two costly maintenance strategies. Reactive maintenance fixes machines only after failure, causing expensive unplanned downtime. Fixed-schedule maintenance services machines on a calendar regardless of actual condition, wasting money on machines that are still healthy. Predictive maintenance — servicing a machine exactly when its sensor data says it needs it — solves both, but normally requires a human to notice an anomaly, then separately estimate its severity, then separately request a repair recommendation.
Solution: UptimeAgent removes those manual hops by exposing three MCP tools designed to be chained autonomously by an LLM agent in a single turn. analyze_sensor_reading compares a machine's recent sensor readings against its own healthy baseline to detect anomalies. predict_failure_window fits a trend line to project remaining useful life in cycles/days, with a confidence score. generate_work_order combines both analyses with a maintenance-manual lookup to produce an actionable ticket: urgency level, issue description, recommended repair action, and estimated remaining life. Asking the agent "How's Engine 3 doing?" triggers all three tools in sequence, unprompted, ending in a finished maintenance ticket.
The Problem: Tons of edible food from restaurants go to waste daily because coordinating pickups with local NGOs is logistically difficult, extremely time-sensitive (perishability), and burdened by language barriers in grassroots charity networks.
What It Does: Built on NitroStack (an MCP framework), this system uses AI to autonomously rescue surplus food. When a restaurant logs a donation, the system queries a PostgreSQL database to calculate Haversine distances to nearby NGOs, strictly filtering by capacity and perishability constraints. It then triggers an automated, real phone call to the optimal NGO using Twilio. Instead of robotic English, it uses Sarvam AI's Text-to-Speech to speak fluently in the NGO's regional language (e.g., Tamil, Hindi). If accepted, it formally allocates the food, assigns the nearest available gig-worker delivery executive to transport it, and handles end-to-end background SMS communications and JSONB audit logging.
Who It Is For:
Restaurants & Donors seeking a zero-friction, automated way to prevent food waste and earn CSR metrics.
Grassroots NGOs & Community Kitchens who can receive localized phone calls instead of struggling with complex English apps.
Delivery Executives looking for optimized, proximity-based gig tasks.
What Makes It Special:
Agentic Autonomy: Acting via Model Context Protocol (MCP) tools, an LLM orchestrator entirely manages dispatching, retry-logic, and negotiations without human input.
Hyper-Local Voice Tech: Integrating Sarvam AI for native-language Twilio voice calls completely removes digital friction for rural or local charity workers.
Safety-First Logistics: Hard-coded SQL geospatial constraints ensure perishable (e.g., non-veg) food is only routed within safe micro-radii (5km).
QuantumSolve AI is a multi-agent AI platform that transforms complex real-world optimization problems into intelligent, explainable solutions. Users describe a problem in natural language, and specialized AI agents collaboratively research the domain, retrieve relevant knowledge, formulate optimization models, evaluate quantum feasibility, benchmark classical and quantum approaches, and generate actionable recommendations. The system is powered by a FastAPI orchestration layer, a deployed NitroStack MCP server with 18 specialized tools, and an interactive Next.js frontend. By combining Retrieval-Augmented Generation (RAG), optimization algorithms, quantum-inspired analysis, benchmarking, and automated report generation, QuantumSolve AI enables organizations, researchers, and developers to solve complex decision-making challenges without requiring expertise in optimization or quantum computing.
Aloka is an AI-native incident resolution platform built entirely around the Model Context Protocol (MCP). By utilizing a network of custom MCP servers (otel-mcp, k8s-test-mcp, git-mcp) alongside a TypeScript LangGraph orchestrator, Aloka empowers specialized AI agents to autonomously diagnose distributed system failures, generate accurate code patches, and verify fixes in real-time.
When someone leaves a team, the reasoning behind their decisions leaves with them — Slack threads and meeting transcripts capture what happened, but never why. Decision Trail is an AI agent that ingests a team's meetings and conversations, reconstructs the rationale behind past decisions, and answers questions like "why did we choose Vendor A over B?" with a structured, sourced answer — decider, date, alternatives considered, and a clickable citation back to the exact moment it came from. If no rationale exists in the record, it says so rather than guessing. Built to make enterprise decision-making auditable, not just searchable.
BidWiserAI is a behavioral protection guard designed specifically for retail investors and active traders prone to making impulsive, emotional decisions in volatile markets. By mapping a user's actual portfolio holdings to real-time market events, the system analyzes trade intents to detect common biases like panic-selling, FOMO, and herd-following. Instead of recommending or blocking trades, it introduces smart, data-backed friction by presenting historical recovery rates and reflective questions before a trade can be executed. What makes it truly unique is its strict adherence to a neutral compliance boundary; the agent never advises you on what to buy or sell, and it programmatically lints its own conversational outputs to guarantee it only provides objective context rather than directional financial advice.
GoAssist AI is an MCP-powered AI assistant that simplifies access to government welfare schemes by checking eligibility, recommending suitable programs, verifying required documents, and guiding citizens through the application process using intelligent multi-agent automation with NitroStack.
Clinical Copilot is an Agentic AI healthcare assistant powered by NitroStack and the Model Context Protocol (MCP). It transforms unstructured medical reports into a structured patient profile and builds a chronological medical timeline. The AI intelligently matches patients with relevant live clinical trials from ClinicalTrials.gov using explainable eligibility reasoning. It then generates physician-ready referral reports to streamline patient enrollment and clinical decision-making. By combining MCP, AI reasoning, and real-world healthcare data, Clinical Copilot helps improve patient care and access to advanced treatments.
Fin-Copilot is an AI financial copilot that reasons through your bills and budgets against your own rules, and only acts after you approve — combining intelligent prediction with human-controlled execution.
**CareBridge AI** is a smart medical assistant powered by **10 AI agents** built on NitroStack MCP. It provides instant symptom checking, lab report analysis, emergency first aid, and personalized health guidance through a sleek glassmorphism interface. The system bridges the gap between patients and healthcare by offering immediate, AI-driven medical insights before visiting a doctor. Built with Node.js backend and deployed on NitroStack Cloud + Render. **Live Demo:** [Your Render URL]
Learn2Earn AI turns learning into a game with real stakes. Pick any topic — or the full JEE syllabus — and it breaks it into 10–16 bite-sized concepts, each with its own lesson and quiz.
Free mode: learn and quiz at your own pace, no money involved.
Paid mode: deposit an amount upfront. It's split equally across every topic and locked. Score 80% or higher on a topic's quiz, and that topic's share unlocks into your real wallet — proof you actually learned it, not just clicked through it. Fall short, and it stays locked until you review and retry.
Who it's for: students prepping for competitive exams, self-learners tackling a new subject, or anyone who studies better with skin in the game.
What makes it special: it's not a chatbot glued onto a syllabus — it's a full learn → prove it → get paid loop, built as a real MCP server so any AI agent can plug straight in. And it never goes dark: even without an API key, it falls back to solid, ready-made lessons and quizzes, so the demo holds up under pressure.
NeuroTwin Evolver is a self-evolving industrial meta-agent built on the NitroStack MCP framework, designed for Industry 4.0/5.0 environments where a fixed automation policy breaks down the moment reality diverges from its training — a heatwave draining UAV batteries faster than expected, a sensor going noisy mid-flight, a carbon-tax policy shifting overnight. Rather than one model wearing every hat, the system is a genuine multi-agent pipeline: a Resource Manager Agent computes a fleet-wide directive by treating energy price, carbon intensity, and delivery pressure as one problem; a Logic-Refactor Agent proposes candidate operational-logic mutations for the affected unit under that directive; a NeuroTwin Validator Agent checks every proposal against symbolic safety rules (battery floor, safety margin, latency ceiling) before anything is allowed near the physical fleet; and a Protocol-Evolver Agent independently monitors swarm-wide communication health, re-engineering the network topology between leader-follower, mesh, and cellular structures when links degrade. The result is a fleet of UAVs, AGVs, conveyor cells, battery racks, and safety gates that detects, proposes, validates, and deploys its own fixes live — visualized through a real-time digital-twin map, a mutation-lab dashboard showing the full agent trail, and per-unit telemetry, all runnable and testable through NitroStudio.
One unhappy customer, three disconnected tools, and no clear owner. Escalation Triage connects every signal, detects risk early, assigns the right person, and keeps the customer informed before frustration turns into churn.
Vulnix is an AI-powered, defensive security assistant built to secure web applications and codebases directly from the developer's workspace. Operating as a lightweight, low-friction Model Context Protocol (MCP) server engineered with the "NitroStack TypeScript SDK", Vulnix bridges the gap between static analysis, endpoint monitoring, and AI-assisted remediation.
Core Capabilities:-
Vulnix exposes granular auditing and remediation tools directly to your AI agents through two primary vectors:
1. Passive Endpoint Audits
Instantly probes target server response headers (HSTS, CSP, X-Frame-Options, X-Content-Type-Options) to evaluate compliance, identify missing configurations, and map surface-level security flaws.
2. Context-Aware Code Remediation
Scans source code for critical security bugs (such as SQL injection vectors, unsafe execution contexts, and hardcoded secrets) and generates drop-in, type-safe, parameterized alternatives.
Who is it for?
* Software Engineers: Fix code vulnerabilities inside your IDE before checking in code, keeping your development loop unbroken.
* AppSec & DevSecOps Teams :Enforce baseline security posture and audit endpoints continuously without writing manual testing pipelines.
* Engineering Leaders: Scale secure-by-default practices across distributed teams by leveraging agentic automation.
What Makes Vulnix Special ?
Traditional application security forces developers to navigate disconnected, heavy scanning dashboards. Vulnix shifts security completely left` by integrating directly with modern AI-native clients and IDEs (such as Claude Desktop, Cursor, and ChatGPT) via the MCP standard.
By empowering agentic AI in the development loop, Vulnix doesn't just flag security issues—it enables agents to autonomously audit endpoints, diagnose vulnerabilities, and commit precise, context-aware code patches directly to your workspace. It's security analysis, automated remediation, and code generation, combined into a single protocol.
An intelligent AI scheduler that helps users organize their daily activities by automatically assigning suitable time slots in Google Calendar. The system identifies scheduling conflicts and reschedules tasks efficiently while maintaining priorities. Developed with NitroStack to provide a seamless and automated time management experience.
Ztockza Copilot is an AI-powered quantitative trading terminal and conversational copilot built on NitroStack MCP. It connects live financial data, multi-factor technical analysis, and automated pattern recognition directly into an interactive dark-mode interface.
What it does:
• Live Candlestick Charting & Target Overlays: Fetches real-time Yahoo Finance OHLC data and renders dynamic support, resistance, and AI target price lines directly on visual charts.
• Multi-Factor Analysis Engine: Combines key price levels, technical momentum indicators, and market psychology (Fear & Greed Index) into actionable trading signals.
• Automated Pattern & Trendline Recognition: Scans historical price bars to detect Ascending Support, Descending Resistance, and Consolidation patterns with pivot-cluster strength scores (0–100).
• Conversational MCP Assistant: Features an embedded chat panel where traders can query stock metrics, resolve stock symbols, and receive instant AI pattern breakdowns with confidence ratings.
Who it is for:
Designed for retail traders, quantitative analysts, and active investors who want institutional-grade technical analysis and chart pattern detection without juggling multiple trading tools.
What makes it special:
Unlike traditional static charts or generic LLM chatbots, Ztockza Copilot acts as a context-aware visual quant assistant. By leveraging custom NitroStack MCP tools, it seamlessly feeds live market math, pivot-cluster scoring, and visual overlay coordinates to the AI—translating raw market noise into actionable, high-confidence trading insights in real time.
AI Placement Guidance is an intelligent career assistant that helps students prepare for campus placements. It analyzes their skills, coding performance, aptitude scores, and mock interview results to identify strengths and weaknesses. Based on their target job role, it recommends personalized learning paths, practice questions, certifications, projects, and interview preparation strategies, enabling students to improve efficiently and increase their chances of getting placed.
ForgeMind is an AI-powered predictive maintenance platform built natively on NitroStack MCP that helps manufacturers detect, investigate, and resolve machine failures faster. Instead of manually searching through machine logs, SOPs, maintenance history, and inventory systems, ForgeMind orchestrates specialized AI agents through standardized MCP tools to automate the entire workflow. The platform features a live Command Center for real-time machine monitoring, an interactive Fault Injection Lab for safe industrial simulations, RAG-powered SOP retrieval, machine history analysis, inventory verification, production impact estimation, and automated work-order generation. Every investigation is streamed live through the NitroStack MCP Gateway, making AI decisions transparent and explainable. ForgeMind enables faster diagnostics, reduced downtime, and smarter maintenance decisions—all from a single intelligent platform.
Enterprise cloud environments leak money and expose attack surface quietly. GPU nodes idle at 2% CPU while billing $3/hour. An S3 bucket sits publicly readable because someone loosened an ACL during debugging and never reverted it. Finding these means manually correlating inventory, utilization telemetry, security findings, and the org directory that tells you who to actually contact.
CloudGuard AI turns that recurring audit into a conversation. It's an MCP server exposing 10 tools that let any LLM reason over live infrastructure data — correlating hourly cost against CPU utilization to compute exact monthly waste, auditing findings by severity, resolving resources to named owners with escalation paths, and generating Terraform remediation.
It files real GitHub P1 tickets, posts Slack Block Kit incident cards with approve/escalate buttons, and closes issues once remediated.
The agent has no write access to infrastructure — not by permission, but because those tools don't exist in the schema.
PharmaGuard is an AI-powered MCP server that acts as a real-time safety net between the doctor's prescription intent and the patient. It intercepts natural language prescription requests, runs several parallel safety checks across multiple clinical dimensions, and delivers a color-coded risk assessment with actionable alternatives.
Continuum Forge bridges the generational skill gap in industrial manufacturing by turning fragile, unwritten expert knowledge into verifiable, automated safety rules.
WHAT IT DOES:
Over 80% of critical manufacturing rules of thumb live only in senior technicians' heads. Continuum Forge operates an Agentic Model Context Protocol (MCP) pipeline that:
1. Grounded Elicitation: Ingests raw interview transcripts from senior technicians.
2. Rule Codification: Codifies unwritten human heuristics into machine-readable Structured JSON AST (Abstract Syntax Tree) rules.
3. Database Validation: Validates rules statistically against real-time & historical Neon PostgreSQL sensor telemetry (e.g. Pump B vibration & temperature logs).
4. Real-Time Coaching: Delivers instant guidance to junior operators with dynamic verbosity settings ("short" for emergency fixes vs "detailed" for mentor training).
5. Langfuse Observability: Wraps every tool call in Langfuse telemetry for 100% auditability and zero AI hallucination.
WHO IT IS FOR:
Smart factories, industrial plant managers, and junior technicians looking to preserve senior expertise before retirement and prevent catastrophic equipment burnouts.
WHAT MAKES IT SPECIAL:
Built natively on the NitroStack MCP Framework with custom Next.js MCP Widgets (Rule AST Visualizer, Emergency Guidance Card, DB Visualizer), live Neon DB telemetry, and full Langfuse tracing.
NutriKids is an Agentic AI-powered pediatric nutrition intelligence platform built using NitroStack's Model Context Protocol (MCP). Parents often struggle to obtain reliable, personalised guidance for concerns such as growth assessment, nutritional deficiencies, meal planning, and child health. Traditional AI chatbots typically rely on a single prompt, making it difficult to coordinate specialised healthcare knowledge across multiple domains.
To address this, NutriKids introduces a Supervisor Agent that analyses each user query and dynamically orchestrates specialised MCPs based on the required expertise. The platform currently includes three MCPs: Growth MCP for WHO-based growth assessment, Medical MCP for symptom and nutritional deficiency analysis, and Food MCP for personalised meal planning and dietary recommendations. Each MCP independently performs its specialised reasoning, while the Supervisor Agent reconciles their outputs into a single structured, parent-friendly response.
Built with NitroStack, the solution demonstrates a modular and scalable Agentic AI architecture where new healthcare capabilities can be added as independent MCPs without changing the existing system. By combining intelligent orchestration, evidence-based recommendations, and personalised nutrition guidance, NutriKids showcases how MCP can be applied to solve real-world healthcare challenges through collaborative AI agents rather than a single monolithic model.
video - https://canva.link/8xmwfwfm6xp3fmg
Interactive Quiz:
An AI-powered MCP application that dynamically generates interactive multiple-choice quizzes on any topic and difficulty using a language model. It provides instant feedback, explanations, and a final score directly within the ChatGPT interface.
Mind Map Generator MCP:
is an MCP (Model Context Protocol) server that generates interactive visual mind maps from a user's topic or prompt. It automatically organises concepts into a clear hierarchical structure, helping users understand, revise, and explore complex subjects more effectively.
What is the Amrita University Coimbatore NitroStack hackathon?
The Amrita University Coimbatore hackathon is a NitroStack × MCP To The Moon buildathon where student teams design, build, and deploy MCP (Model Context Protocol) apps and servers. Every project on this page is a real, submitted MCP application built by a student team and deployed on NitroStack — the full-stack platform for building and shipping agentic AI and MCP apps.
What are MCP apps and MCP servers?
MCP apps and MCP servers are applications built on the Model Context Protocol (MCP) — an open standard that lets AI agents securely connect to tools, data, and APIs. An MCP server exposes tools, resources, and prompts that any MCP-compatible AI agent can call, turning a large language model into an agentic AI system that can take real actions. NitroStack lets you build, deploy, and scale these MCP apps end to end.
How many projects were built at Amrita University Coimbatore?
183 MCP projects were submitted at the Amrita University Coimbatore hackathon. You can browse each one below, watch the demo, read the write-up, and open the source code and live MCP endpoint.
How can I build my own MCP app?
You can build your own MCP app on NitroStack (nitrostack.ai) — it provides the SDK, cloud deployment, and studio to go from idea to a deployed, agentic AI MCP server. Explore the docs at docs.nitrostack.ai and browse community builds on r/mcptothemoon.