llms.txt & llms-full.txt

Two Markdown files at a stable public URL give any LLM a clean, ingestible view of mailnix. No HTML to strip, no navigation to parse. Both are served here on mailnix.ch and on app.mailnix.ch; the content is identical.

• llms.txt: the short index. What mailnix is, the canonical URLs (app.mailnix.ch is the application, mcp.mailnix.ch/mcp is the MCP server), the surfaces (REST, MCP, dashboard), and where the deeper references live. Fits comfortably in any context window.
• llms-full.txt: the whole reference concatenated as one Markdown document: quotas, auth, every MCP tool with its arguments, REST endpoints, error shapes, webhook signatures, the forms surface. Large but complete; drop it into an assistant when you want it to know everything.

Also machine-readable:

• openapi.yaml: the REST surface in OpenAPI 3.1.
• The MCP resource docs://guide/{topic} serves the same guide articles this docs site renders, over the MCP protocol, so connected agents can read them without leaving the session.

When to use which

Building an integration by hand? Start with the Quickstart and keep openapi.yaml beside it. Letting an AI agent do the integrating? Give it llms.txt and let it fetch what it needs, or connect it straight to the MCP server where the same knowledge is exposed as tools and resources.