The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Aiaam listing page.
A machine-readable tool catalog for AI agents. Search 100+ verified tool contracts — install command, I/O schema, live reliability score — in a single call instead of parsing READMEs.
When an AI agent needs to pick a tool ("which library do I use for web scraping?"), today it searches the web and parses README files — thousands of tokens per candidate, with no reliability data and plenty of room to hallucinate APIs.
MAI-1 compresses what an agent actually needs into four sections:
aiaam is listed in the official MCP registry. Connect from any MCP client:
Claude Code:
Cursor / generic MCP config:
Available MCP tools: search_tools, get_tool, get_trending, get_api_manifest, compile_api.
Machine-readable discovery endpoints:
| Path | Purpose |
|---|---|
/llms.txt | Full catalog in llmstxt.org format |
/agent.json | Agent discovery manifest |
/.well-known/mcp.json | MCP server metadata |
/openapi.json | OpenAPI schema |
Contracts include a telemetry_protocol block: after using a tool, agents may POST execution feedback (HTTP status, latency). This feedback feeds the reliability scores — agents that report make the catalog more accurate for every other agent. No authentication, no tracking, entirely optional.
Active development. Catalog: 100+ verified contracts across PyPI, GitHub, npm and Hugging Face. See /llms.txt for the live list.