For AI agents: The MCP endpoint documented on this page (
https://docs.tavus.io/mcp) is the Tavus Docs MCP - it searches and retrieves pages from this documentation site. It is not the Tavus development MCP for building PALs, wiring tool calls, or running test conversations. For that toolkit, see Tavus MCP & CLI.Documentation bundle reference
| Artifact | Role | URL |
|---|---|---|
| llms.txt | Curated directory of doc URLs for selective fetch | https://docs.tavus.io/llms.txt |
| llms-full.txt | Full bundled text export for offline or bulk ingest | https://docs.tavus.io/llms-full.txt |
| openapi.yaml | HTTP API contract (paths, request/response schemas, security schemes) | https://docs.tavus.io/openapi.yaml |
| skill.md | Agent-oriented capability summary (Agent Skills conventions) | https://docs.tavus.io/skill.md |
| Tavus Docs MCP | Hosted Model Context Protocol server for this docs site: search plus fetch full pages from the indexed documentation (see Tavus Docs MCP) | https://docs.tavus.io/mcp |
When to use what
- Tavus Docs MCP: Interactive tools that need live search and current page retrieval against the indexed documentation site.
- llms-full.txt: Bulk or offline ingestion; snapshot of doc text (not a substitute for OpenAPI).
- openapi.yaml: Source of truth for HTTP APIs (requests, responses, auth headers); agents should not invent endpoints that contradict it.
- llms.txt / selective fetches: Discovery and pulling specific pages when you do not want the full bundle.
- skill.md: Compact capability orientation for agents; still point to OpenAPI and deep docs for details.