# AIEndpoint

> Open standard for AI agent service discovery. Like robots.txt told crawlers where not to go, /ai tells AI agents what your service can do.

AIEndpoint defines a convention: any web service places a JSON endpoint at `yoursite.com/ai` describing its capabilities, authentication, and rate limits — in ~800 tokens instead of 10,000–50,000 from scraping HTML.

The registry at aiendpoint.dev indexes services implementing the standard, and an MCP server auto-discovers capabilities for any website.

## Docs

- [Spec v1.0](https://aiendpoint.dev/docs/spec): The /ai endpoint specification
- [Quick Start](https://aiendpoint.dev/docs/quick-start): Implement /ai in 10 minutes
- [Concepts](https://aiendpoint.dev/docs/concepts): Token efficiency, capability design, auth model, versioning
- [MCP Server](https://aiendpoint.dev/docs/mcp-server): Auto-discover any website from Claude, Cursor, or Claude Code
- [Validation](https://aiendpoint.dev/docs/validation): How the 0–100 compliance scoring works
- [Changelog](https://aiendpoint.dev/docs/changelog): Release history

## API

- [Registry API](https://api.aiendpoint.dev/ai): The registry's own /ai endpoint (JSON)
- [Search services](https://api.aiendpoint.dev/api/services): Browse registered services
- [Validate](https://api.aiendpoint.dev/api/validate?url=https://api.aiendpoint.dev): Check /ai compliance
- [Community specs](https://api.aiendpoint.dev/api/community/): Lookup community-generated specs

## Links

- [GitHub](https://github.com/aiendpoint/platform): Source code (Apache 2.0)
- [npm](https://www.npmjs.com/package/@aiendpoint/mcp-server): MCP server package
- [Contributing](https://github.com/aiendpoint/platform/blob/main/CONTRIBUTING.md): How to contribute

## Optional

- [Roadmap](https://aiendpoint.dev/docs/roadmap): What's next
- [Why](https://aiendpoint.dev/why): Why we built this