MCP Discovery

Before an AI agent can use lnkify, it needs to discover what tools and resources are available. lnkify supports three discovery mechanisms.

llms.txt

A machine-readable discovery file at https://lnkify.io/llms.txt:

• Content-Type: text/markdown
• Caching: 1 hour
• Contains: MCP endpoint URL, tool names with short descriptions, GraphQL endpoint, links to full docs

Agents fetch this file to learn the basic shape of the lnkify MCP server without pulling in full schemas.

llms-full.txt

An expanded reference at https://lnkify.io/llms-full.txt:

• Content-Type: text/markdown
• Caching: 1 hour
• Contains: Full tool schemas with field types and descriptions, argument shapes, worked examples

This is ideal when the agent needs parameter-level detail to construct tool calls correctly.

How Agents Use These Files

1. Fetch llms.txt to discover the server endpoint and tool names.
2. Follow the embedded links to https://docs.lnkify.io/developer/mcp/ for human-readable docs.
3. Optionally fetch llms-full.txt for detailed schemas.
4. Connect to the MCP server and use tools/list for runtime discovery.

Runtime Discovery

In addition to the static files above, the MCP server supports standard runtime discovery methods:

tools/list

{
  "jsonrpc": "2.0",
  "method": "tools/list",
  "id": 1
}

Returns every registered tool with its name, description, and JSON Schema for inputs. This is the canonical way to discover tools at runtime.

resources/list

{
  "jsonrpc": "2.0",
  "method": "resources/list",
  "id": 2
}

Returns every registered resource with its URI, name, description, and MIME type.

File Generation

Both llms.txt and llms-full.txt are generated from the tool registry at build time — the same data that powers tools/list. The generation logic lives in server/src/mcp/llms.ts.

Human-Readable Docs

The full docs site at https://docs.lnkify.io/developer/mcp/ is the human-readable counterpart to llms.txt. Use it for integration planning and understanding the MCP server's full capabilities.