Open Innovation Code for changeSubmitted July 26, 2026

WARDEN — an MCP server that assumes every website it reads is trying to trick it

An MCP app on the Model Context Protocol built by Code for change at the Amrita University Coimbatore NitroStack × MCP To The Moon hackathon and deployed on NitroStack.

About this project

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.

Open Innovation track

Solve any real-world problem with AI, regardless of industry or domain.

Source code

Team Code for change

  • S AshvinsaiLead

  • Ragav U

  • Adhi Arunasalam. A

  • P K PRANAV

← All Amrita University Coimbatore projects

Frequently asked questions

What does WARDEN — an MCP server that assumes every website it reads is trying to trick it do?
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.
Who built WARDEN — an MCP server that assumes every website it reads is trying to trick it?
WARDEN — an MCP server that assumes every website it reads is trying to trick it was built by team Code for change at the Amrita University Coimbatore NitroStack × MCP To The Moon hackathon, in the Open Innovation 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.