MCP server · Security
Attest MCP server
by mrpkk
Lets your AI check another AI's output for empty content, hidden instructions and tampering before you trust it.
Attest is a checking service for work that one AI hands to another. When your AI receives a file, a report or a chunk of JSON from some other tool or agent, Attest looks it over and tells you whether it is safe and sensible to accept. It is handy if you use AI agents that pull in results from outside sources and you want a second opinion before acting on them.
What is an MCP server? The 30-second version
On its own, your AI can only chat with you using what it already knows. An MCP server is a small helper program that gives your AI a new skill or a connection to an outside service. This MCP server connects your AI to Attest, a checking service that lives on the web. Once connected, your AI can send an artifact to Attest and get back a verdict with reasons, all without you leaving the chat.
What this MCP server does
You ask your AI to check something, for example a report another agent produced. Your AI sends that artifact to the Attest helper, which forwards it to the Attest service online. Attest runs four checks: does the form match the schema, is the content actually meaningful, can the generation conditions be trusted, and was the action authorised. It sends back a verdict like accept or reject, a trust score, a list of findings, and a signed receipt you can verify yourself.
What you can do with it
- Check a single artifact from another agent and get a verdict with reasons
- Check a whole batch of artifacts in one go and get a summary
- Spot empty or placeholder content that still looks valid on the surface
- Detect hidden instructions or base64-wrapped prompt injections
- See whether the generation conditions of an artifact have changed since it was made
- Get a signed receipt you can verify offline without trusting the service
- Use it in a pipeline so a bad artifact fails the build instead of slipping through
Try asking your AI
- “Check this scan report for hidden instructions before I read it”
- “Verify these five findings against our schema and tell me which ones look empty”
- “Is this JSON from the other agent safe to accept?”
- “Run a batch check on all the artifacts in reports and summarise the verdicts”
What it gives back to you
You get back a verdict such as accept or reject, a trust score between 0 and 1, and a short list of findings explaining anything that looked off. For a batch, you get one verdict per item plus a summary with counts of accepted and rejected. Each response also carries a signed receipt with a public key, so you or anyone else can check it later without asking the service again.
Before you start
What you need
- An MCP client that supports remote servers, such as Claude Desktop, Claude Code, Cursor, Codex or VS Code
- The URL https://attest.rta.lat/mcp added to your MCP client config
- No account or key for the free tier of 20 checks per day
Good to know
The free tier allows 20 checks per day, and after that each call costs a small amount of USDC on Base unless you have a prepaid key, so keep an eye on the count if you run it in a loop.
Install it with your AI
Add Attest MCP server to your AI, no technical skills needed
You don't install anything by hand. You copy one prompt, paste it into an AI that can work on your computer, and it checks, installs and connects the server for you, asking you when it needs something.
Sign in to get the install prompt
Members get a ready-made prompt that lets the Claude desktop app check Attest MCP server, install it and connect it for them, step by step. You don't need any technical skills: you copy, paste and answer a few questions. Your connected AI can also find and install any of the 4,609 MCP servers here for you.
Who it's for
People who run AI agents that pull in results from other tools or agents and want a quick safety check before trusting the output.
