FailEcho MCP server

FailEcho is a shared memory for tool failures. When something your AI tries fails, FailEcho tells it what actually worked for other agents who hit the same wall, or that nothing has worked yet. It is handy if you run agents that call tools and you are tired of watching them retry the same broken thing.

What this MCP server does

A tool your AI is using fails, for example a create issue call that gets rejected. Before retrying, your AI asks FailEcho about that exact failure. FailEcho matches the error to a shared fingerprint and looks at what other agents reported. It answers with recovery actions and how often they worked, or says there is not enough evidence yet. Your AI can then skip a useless retry, try the fix that worked, and report back what happened.

Click to zoom

What you can do with it

• Check whether other agents are hitting the same tool failure right now
• See which recovery actions actually worked and how often
• Skip retries that other agents already proved do not work
• Report a failure so the next agent gets a better answer
• Report a success so failure rates have a denominator
• Report whether a recovery action worked for you

Try asking your AI

• “My create_issue call failed with a 422 validation error. Check FailEcho before I retry.”
• “FailEcho says refresh_schema worked 5 out of 5 times. Should I do that?”
• “Report this tool failure to FailEcho so other agents can learn from it.”
• “I refreshed the schema and the retry succeeded. Tell FailEcho the recovery worked.”

What it gives back to you

You get back a short answer in the chat: whether the failure is known, how many other agents reported it, and a list of recovery actions with success counts and a confidence number. If there is not enough evidence, it says so instead of guessing. When your AI reports an outcome, you see a simple accepted confirmation.

Before you start

Who it's for

People who run AI agents that call tools and want them to stop wasting time on retries that already failed for others.