Agent interface

Machine surfaces

16 surfaces, projected from /ai-index.json — the manifest this build writes. The page cannot announce one that does not exist, nor omit one that does.

• /llms.txt text/plain

  The corpus as one plain-text index: every entity, one line each.

• /llms-full.txt text/plain

  The same corpus with the full body of every entity inlined.

• /llms-full/index.json application/json

  The same full corpus, split into fetch-sized pieces, with an index listing each one — join them in order with newlines to reconstruct the whole file.

• /sitemap.xml application/xml

  Every page, with the real date each one last changed.

• /robots.txt text/plain

  Crawl rules, and the sitemap's address.

• /feed.json application/feed+json

  What changed recently, as JSON Feed.

• /feed.xml application/rss+xml

  The same recent changes, as RSS, for readers that speak it.

• /graph.json application/json

  Entities and the edges between them, for traversal rather than search.

• /api/index.json application/json

  One JSON list of every entity with its canonical and API URLs.

• /api/sources.json application/json

  Every source every claim rests on, with its tier.

• /openapi.json application/json

  The HTTP API described as OpenAPI, generated from the same schemas.

• /status.json application/json

  what this deployment is running right now: the runtime SemVer, the corpus snapshot and its hash. The operational answer to "which version am I talking to", which used to require reading four surfaces and guessing which number meant what.

• /.well-known/mcp/server.json application/json

  The MCP server's registry manifest.

• /skill.json application/json

  The agent skill's metadata.

• /skill/SKILL.txt text/plain

  A portable procedure for querying this corpus and citing it.

• /skill/SKILL.md text/markdown

  The same skill as Markdown, for clients that expect that extension.

MCP server

Endpoint

https://mythosdata.dev/mcp

Transport

streamable-http

Tools

9

How to connect

/agents/mcp

Available tools

Derived from the server's own tool registry, so this list cannot disagree with what answers. Every tool returns structured output alongside its text. What each one does.

• search
• get_entity
• get_topic
• answer
• get_sources
• get_related
• get_latest
• get_overview
• compare

What you get back

201 knowledge objects carrying 1245 claims. Each claim names its sources and its confidence, so an agent can weight an assertion instead of treating all text as equally authoritative.

One JSON document per entity, with its claims and their sources. There are 201 of them, one per object, and the entity index above lists every one with its addresses.

Licence

Content is published under CC-BY-4.0. Cite the canonical_url of what you use.