Let another AI read the platform documentation
VitePress is the human presentation layer. The Skill, llms.txt, full context, and OpenAPI document are the machine-readable entry points.
| AI capability | Recommended entry |
|---|---|
| Can install Skills | Developer Skill |
| Can read a URL | llms.txt |
| Needs a single full context | llms-full.txt |
| Generates typed API code | openapi.json |
| Reads one page at a time | Quickstart |
Example prompt:
text
Read https://<platform-host>/developers/llms.txt first.
Then integrate Xiaoya Open Platform OAuth 2.0, the available-model catalog,
and the model gateway into this project. Keep Personal and Team tokens isolated,
and never expose a token or client secret to browser code.llms.txt is an AI-friendly documentation convention, not an authorization protocol. The build derives llms-full.txt from the same bilingual Markdown pages, avoiding a second manually maintained knowledge base.
A separate documentation MCP service is not needed for this static first version. Add one only when the corpus becomes large or access must return dynamic tenant-specific information.