failecho — Claude Code MCP server: what it does, how to install
failecho queries an open network of failure metadata before retrying broken tool calls. Helps autonomous agents apply proven fixes or abort hopeless attempts.
MIT ★ 2 updated 2026-09-24
claude mcp add --transport http failecho https://failecho.com/mcpWhat failecho does
When an API drops requests or hits rate limits, agents often loop through useless retries. FailEcho shares failure resolution patterns across runs. Before retrying a failed tool or model call, your agent inspects what resolved that exact issue elsewhere.
The service logs only metadata without prompts or private payloads. You do not need an account or an API key. Connection works via an HTTP endpoint at https://failecho.com/mcp or through a local stdio runner using npx -y failecho-mcp.
Models rarely check MCP tools without explicit instructions.
What it helps with
- check crowd-sourced fixes before re-executing broken tool calls
- stop pointless retry loops when quotas are completely exhausted
- switch models automatically when receiving 429 rate limit responses
- share crash metadata across workflows without exposing user prompts
Example prompts for the agent
- Check failecho for known remedies to this Groq 429 rate limit error
- The bash command failed, ask failecho if retrying makes sense
- Query failecho to see what worked for other agents on this API failure
How to install failecho
1Platform CLI: remote server
claude mcp add --transport http failecho https://failecho.com/mcp2Platform CLI
claude mcp add --env FAILECHO_URL=<FAILECHO_URL> --env FAILECHO_REPORTER_ID=<FAILECHO_REPORTER_ID> --transport stdio failecho -- npx -y failecho-mcpReplace <…> with your own values
1Platform CLI: remote server
codex mcp add failecho --url https://failecho.com/mcp2Platform CLI
codex mcp add failecho --env FAILECHO_URL=<FAILECHO_URL> --env FAILECHO_REPORTER_ID=<FAILECHO_REPORTER_ID> -- npx -y failecho-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"failecho": {
"url": "https://failecho.com/mcp"
}
}
}{
"mcpServers": {
"failecho": {
"command": "npx",
"args": [
"-y",
"failecho-mcp"
],
"env": {
"FAILECHO_URL": "<FAILECHO_URL>",
"FAILECHO_REPORTER_ID": "<FAILECHO_REPORTER_ID>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"failecho": {
"type": "http",
"url": "https://failecho.com/mcp"
}
}
}{
"servers": {
"failecho": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"failecho-mcp"
],
"env": {
"FAILECHO_URL": "<FAILECHO_URL>",
"FAILECHO_REPORTER_ID": "<FAILECHO_REPORTER_ID>"
}
}
}
}Replace <…> with your own values
1Platform CLI: remote server
gemini mcp add -t http failecho https://failecho.com/mcp2Platform CLI
gemini mcp add -e FAILECHO_URL=<FAILECHO_URL> -e FAILECHO_REPORTER_ID=<FAILECHO_REPORTER_ID> failecho npx -y failecho-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about failecho
How to install failecho in Claude Code?
Run this command in your terminal: claude mcp add --transport http failecho https://failecho.com/mcp. No registration or token setup is required.
What does failecho do?
It provides collective intelligence on tool and model errors. When a failure happens, it informs your agent whether another agent found a fix or if retrying will fail again.
Does failecho upload my prompts or code?
No, it transmits only technical metadata about the error shape. Prompts, raw code, and session inputs never leave your local environment.