Use these docs with AI
These docs are machine-readable end to end. There are three ways to work with them from an AI tool, in increasing order of what they can do.
Every page also has a Use with AI button, top-right, which shows the configuration for connecting a coding agent to the MCP server described below.
Copy any page as Markdown
Use the Copy page button at the top of a page, or append .md to any docs URL
to fetch it as clean Markdown:
https://nym.com/docs/developers/mcp (the page)
https://nym.com/docs/developers/mcp.md (the same page as Markdown)Paste it into any chat, or fetch it from a script, to give an assistant the exact page without scraping HTML.
Bulk ingest: llms.txt
For an index of every page, or the full corpus in one file:
https://nym.com/docs/llms.txt (index of pages, for discovery)
https://nym.com/docs/llms-full.txt (every page's content in one file)Point an agent at llms.txt to discover pages, then dereference each entry to its
.md.
MCP server (for coding agents)
This is the capable option. Point your agent at a single endpoint, and it can search the docs and the Nym source, and read live network state, as structured tools. Claude Code, Cursor and others are supported:
https://nym.com/docs/api/mcpSee the MCP server reference for the tool catalogue, client configuration, and examples. The live tools (node counts, gateway health, token supply, chain status) are the reason to use this over a static file: an agent can ask "is gateway X in the active set right now?" and get a current answer.
How all of this works
All three surfaces come from the same build. The MCP server reads a semantic index
constructed when the docs deploy, served as a static file rather than from a vector
database. The Markdown files and llms.txt are written by the same pass over the
same pages.
Every search result carries a deep link to the section it came from. Any claim an agent makes can be checked against the page behind it.
The MCP endpoint is open: no API key, no account. It embeds each incoming query at Nym's expense and returns raw sections without generating anything, so the only tokens you spend are your own agent's. See how search works for the mechanism.