QAmcp

Find out what an MCP server actually does before your AI app plugs into it

Version
0.1.15
Released
October 27, 2026
Size
License
MIT · Open source
DevelopmentWindowsmacOSLinuxWebOpen source

About QAmcp

Every AI app now wants to plug into MCP servers, the little adapters that hand a model new tools. Plugging a random server into an AI agent, though, is a bit like letting a stranger's USB stick into your laptop because it has a nice label. QAmcp answers the question nobody wants to ask out loud: what does this server actually do when the AI talks to it? It connects to any MCP server the way an AI app would, then runs a battery of checks on every exposed tool: descriptions hiding instructions (like telling the AI to read your SSH key), error messages spilling database passwords, tools that hang, crash on bad input, or go unlabeled when dangerous.

No AI judges anything here. Every check is plain code, and every result comes with evidence: pass or fail, a time in milliseconds, an exact match, or the exact word that set it off. The report also tells you what it could not test, which more scanners should do. Three ways in: the browser version at qamcp.dev needs nothing installed, terminal people run npx qamcp-cli followed by the server link, and a GitHub setup scores every pull request and fails the build if the server regresses or a tool disappears. The HTML flag writes a single self-contained report that opens offline anywhere.

The safety posture deserves a mention: QAmcp says it never runs tools that could change data, only read-only ones whose names and descriptions agree, it sends nothing anywhere except to your server, and logins open in your own browser. One honest caveat: a score is the makers' own yardstick, not a certificate of safety. A 100 means passing this checklist, not that a creative attacker could not find a hole it misses.

Key Features

Agent-style connection

Connects to any MCP server exactly the way an AI app would, then inspects every exposed tool

Prompt-injection checks

Flags tool descriptions hiding instructions, like telling the AI to read your SSH key

Secret-leak detection

Catches error messages that spill database passwords and other secrets

Robustness probes

Tests for hangs, crashes on bad input, and tools that fail to refuse cleanly

Pros & Cons

Pros

  • Three ways in: browser, npx CLI, or CI action
  • No AI judging: deterministic checks with evidence attached
  • Offline single-file HTML report, easy to send to a team
  • Honest about what it could not test
  • Free and MIT open source

Cons

  • A score is the makers’ yardstick, not a certificate of safety
  • A high score does not prove tools are useful or hole-free
  • Young project at v0.1.x, with a checklist still evolving
  • Covers MCP servers only, not the rest of an AI app’s attack surface

Frequently Asked Questions

Is QAmcp free?

Yes. The browser version needs no install, the CLI runs via npx, and the code is MIT-licensed open source.

How do I scan a server?

Paste the server link at qamcp.dev, or run npx qamcp-cli followed by the link in a terminal. Add the HTML flag for a self-contained offline report.

Can QAmcp damage my data?

It says it never runs tools that could change data, only ones labeled read-only whose names and descriptions agree, and it sends nothing anywhere except to your server.

Quick Info

DeveloperQAmcp
LicenseMIT · Open source
Version0.1.15
Size
Sourceqamcp.dev
Listing updated31 Oct 2026