llms.txt

The MediaLocker MCP server publishes an llms.txt discovery document — a plain-text, Markdown-formatted summary that lets AI assistants and coding agents learn how to connect and what tools are available without reading the full docs.

Where it's served

The document is served unauthenticated by the MCP server at three paths:

URL
https://mcp.medialocker.io/llms.txt	Primary
https://mcp.medialocker.io/.well-known/llms.txt	Well-known location
https://mcp.medialocker.io/mcp/llms.txt	Under the MCP path

curl https://mcp.medialocker.io/llms.txt

Scope

llms.txt is published by the MCP server only — it documents the MCP surface (connection, auth, tools, rate limits, error codes). The REST API does not serve an llms.txt; for machine-readable REST details use the OpenAPI spec at https://api.medialocker.io/api/openapi.json.

What it contains

The document is a human- and machine-readable Markdown summary covering:

• Server URL & transport — https://mcp.medialocker.io/mcp, Streamable HTTP with Mcp-Session-Id session management.
• Authentication — bearer API keys (and dashboard JWTs), the per-tenant scope/allowlist model, and how to create a key.
• Available tools — grouped by area (search, buckets, objects, tags, sets, usage), including which require write/delete/admin.
• Destructive tools — the firewall-gated delete_object, delete_bucket, and purge.
• Rate limits — 120 requests/minute (burst 120), 50,000/day per tenant.
• JSON-RPC error codes and short request examples.

It is versioned with the MCP server release (a Server version: line at the top), so it stays in step with the deployed tool catalog. For the full reference, see the MCP Server section.