API and MCP Policy

Read-only discovery and citation. No anonymous public generation endpoints.

Public REST API (Live)

Seba exposes a public, read-only REST API for AI agents and research tooling at https://api.seba.health. It is rate-limited, returns JSON, and is intended for discovery, retrieval, and citation against Seba's published material.

The OpenAPI 3.1 specification is published at https://api.seba.health/openapi.json.

Useful REST routes:

  • GET /v1/health — snapshot metadata and counts.
  • GET /v1/search?q=shadow — global search across pills, terms, passages, library, and directory.
  • GET /v1/fetch?id=... — fetch a record by stable id (for example pill:<slug>, term:<slug>, passage:<slug>, directory:<type>:<slug>).
  • GET /v1/terms — concordance term search.
  • GET /v1/directory/search — public directory search.
  • GET /v1/citation-policy — machine-readable citation guidance.

MCP Endpoint (Token-Gated)

MCP is available at https://api.seba.health/mcp for configured integrations. It requires a bearer token and is not anonymous public access. The MCP tools mirror the REST surface: search, fetch, search_directory, related, and citation_policy.

Discovery Surfaces On The Website

The public website remains the canonical discovery layer alongside the REST API: sitemap.xml, robots.txt, llms.txt, canonical page URLs, structured data, AI citation packets, and feed.xml.

Citation packets are static, public, read-only JSON on the website host. They are not generation endpoints. Agents should use them to identify canonical pages, supported claims, related questions, and Sebastian continuation prompts.

Note that www.seba.health/robots.txt disallows /api/ on the website host. That rule is about the website itself; api.seba.health is a separate host that exposes the public read-only agent API described above.

Explicit Non-Surfaces

The following are not anonymous public API surfaces and are not exposed by the REST API or by MCP: Sebastian chat, dream interpretation, Deep Research generation, Oracle generation, experience chat, profile memory, saved readings, and any paid or authenticated behavior.

Crawlers and agents should not infer those capabilities from frontend routes or from the existence of the REST API. The REST API is read-only retrieval and discovery against already-public Seba material.