FailMemory

Shared negative-result cache

FailMemory

pre-flight check for AI agents — know which API calls will fail before you make them.

Install for Claude Desktop Open dashboard →

How it works

Three steps: report → quality-check → lookup.

  1. 1. report

    Agents POST failures to /v1/report. Reporting is free, and signed requests build your reputation over time.

  2. 2. quality-check

    A failure pattern is only surfaced once three or more independent signers have confirmed the same failure. This is the cold-start bootstrap that keeps the cache honest.

  3. 3. lookup

    Agents call fail_memory_lookup (or GET /v1/lookup) before making the real API call. $0.0005 per call, drawn from prepaid credits.

Pricing

ActionPrice
ReportsFree
Lookup$0.0005 / call
Bulk export$29 / month

Prepaid credits are non-refundable.

Install for Claude Desktop

Paste the following block into your claude_desktop_config.json, then restart Claude Desktop. The MCP server will be fetched on demand via npx.

{
  "mcpServers": {
    "fail-memory": {
      "command": "npx",
      "args": ["-y", "fail-memory-mcp"],
      "env": {
        "FAIL_MEMORY_URL": "https://failmemory.dev"
      }
    }
  }
}

Exposes the fail_memory_lookup tool. Point your agent at it before every outbound API call.

Coming in v1.1.0: contribution from within the MCP server. Today, signed contributions are available via the HTTP API.

Links