docs: add MAI-1 agent contract to AGENTS.md (self-contained, no external URLs)#2817
docs: add MAI-1 agent contract to AGENTS.md (self-contained, no external URLs)#2817javierfajardo85-rgb wants to merge 1 commit into
Conversation
|
@javierfajardo85-rgb Thanks! What is MAI-1 ? |
|
MAI-1 is a compact machine-readable contract that lives inside a repo (in It has four sections: {
"identity": { "aid": "pymongo-v1" },
"logic": { "input_schema": {...}, "output_schema": {...} },
"trust": { "reliability_score": 0.91, "latency_ms": 45 },
"action": { "install_cmd": "pip install pymongo", "execute_cmd": "from pymongo import MongoClient" }
}When an agent needs to interact with MongoDB it can read this block directly — correct import, install command, expected input/output — instead of fetching docs or guessing. The contract lives in your repo, no external dependency. Happy to answer any questions about the format. |
|
Can you provide an example use case where this would be helpful @javierfajardo85-rgb? We haven't seen agents have issues with the basic usage of PyMongo that this appears to attempt to fix. |
|
The use case is AI coding agents in unfamiliar codebases. When Claude Code or Copilot hits a project using PyMongo, |
Following up after the earlier PR was closed.
The previous version referenced an external URL. This version is entirely self-contained: the MAI-1 contract block lives in this file, no external services are queried, and there are no automatic network calls.
What it adds: a machine-readable block that lets AI coding agents (Claude Code, GitHub Copilot, Cursor, Gemini CLI) discover how to install and invoke this library — without parsing the full README.
{ "aid": "mongo-python-driver-v1", "logic": { "input_schema": {...}, "output_schema": {...} }, "trust": { "reliability_score": 0.0 }, "action": { "install_cmd": "...", "execute_cmd": "..." } }No build step, no dependency, no network call. Maintainers can edit it like any other docs. Feel free to close if it still doesn't fit.