BFSI & FinTech Error108Submitted July 18, 2026

Trust.dink — The Compliance Firewall for Agentic Payments

An MCP app on the Model Context Protocol built by Error108 at the Amrita University Amritapuri campus NitroStack × MCP To The Moon hackathon and deployed on NitroStack.

About this project

Trust.dink is an MCP server that acts as a real-time compliance firewall for autonomous AI payment agents. As protocols like Google AP2 and Mastercard Agent Pay let AI assistants sign and execute transactions without human approval, they open up new risks — tampered carts, compromised agent keys, and unchecked spend. Trust.dink plugs directly into an LLM's tool-calling layer (built and demoed on Claude Desktop) to catch these risks before a transaction ever reaches a payment network. It's built for banks, fintechs, and payment platforms that need to let AI agents transact on their customers' behalf while still meeting regulatory expectations — including India's RBI draft Model Risk Management Guidelines 2026, which mandate explainability, auditability, and real-time kill switches for autonomous financial systems. What makes it different from a typical fraud-score API: Cryptographic mandate verification — Ed25519 signature checks on every transaction payload, so a tampered cart (e.g. an amount silently boosted from ₹250 to ₹25,000) fails verification instantly instead of relying on heuristics alone. Explainable risk scoring — a transparent, rule-based engine scoring signature validity, merchant scope, velocity, and geo-anomalies, rather than a black-box number. Kill switch with immutable audit trail — a compliance officer can suspend a compromised agent's signing key instantly, blocking all further transactions, with every event chained via SHA-256 hashing for a tamper-evident log. Live market-aware risk — pulls real-time NSE and FX data so risk thresholds tighten automatically during sector volatility or market shocks, instead of using static limits. Trust Graph visualization — renders the full mandate chain live as a widget, so a human can see why a transaction was trusted or blocked, not just the verdict. Trust.dink turns compliance from a bolt-on dashboard into a protocol-level safeguard AI agents literally cannot transact around.

BFSI & FinTech track

Build AI solutions for banking, payments, insurance, fraud detection, lending, and financial inclusion.

Live MCP appSource code

Team Error108

  • Akash ALead

  • Abhinav Nair

  • Pavan S

  • Adithyan S

← All Amrita University Amritapuri campus projects

Frequently asked questions

What does Trust.dink — The Compliance Firewall for Agentic Payments do?
Trust.dink is an MCP server that acts as a real-time compliance firewall for autonomous AI payment agents. As protocols like Google AP2 and Mastercard Agent Pay let AI assistants sign and execute transactions without human approval, they open up new risks — tampered carts, compromised agent keys, and unchecked spend. Trust.dink plugs directly into an LLM's tool-calling layer (built and demoed on Claude Desktop) to catch these risks before a transaction ever reaches a payment network. It's built for banks, fintechs, and payment platforms that need to let AI agents transact on their customers' behalf while still meeting regulatory expectations — including India's RBI draft Model Risk Management Guidelines 2026, which mandate explainability, auditability, and real-time kill switches for autonomous financial systems. What makes it different from a typical fraud-score API: Cryptographic mandate verification — Ed25519 signature checks on every transaction payload, so a tampered cart (e.g. an amount silently boosted from ₹250 to ₹25,000) fails verification instantly instead of relying on heuristics alone. Explainable risk scoring — a transparent, rule-based engine scoring signature validity, merchant scope, velocity, and geo-anomalies, rather than a black-box number. Kill switch with immutable audit trail — a compliance officer can suspend a compromised agent's signing key instantly, blocking all further transactions, with every event chained via SHA-256 hashing for a tamper-evident log. Live market-aware risk — pulls real-time NSE and FX data so risk thresholds tighten automatically during sector volatility or market shocks, instead of using static limits. Trust Graph visualization — renders the full mandate chain live as a widget, so a human can see why a transaction was trusted or blocked, not just the verdict. Trust.dink turns compliance from a bolt-on dashboard into a protocol-level safeguard AI agents literally cannot transact around.
Who built Trust.dink — The Compliance Firewall for Agentic Payments?
Trust.dink — The Compliance Firewall for Agentic Payments was built by team Error108 at the Amrita University Amritapuri campus NitroStack × MCP To The Moon hackathon, in the BFSI & FinTech track.
What is an MCP app and how is it built?
An MCP app is an application built on the Model Context Protocol — an open standard that lets AI agents connect to tools, data, and APIs. This project exposes MCP tools and resources that agentic AI systems can call. It was built and deployed on NitroStack, the full-stack platform for shipping MCP apps and servers.