FousDocs
Studio

Use with coding agents

Point Claude Code, Codex or Cursor at these docs, and give them the Fous MCP server.

These docs are written for people and for the agents that write code with them. Every page is plain Markdown on request, and the whole site fits in one file.

Markdown for every page

WhatURL
One page as MarkdownAdd .md to its URL, for example https://docs.fous.com/quickstart.md.
The same, by content typeSend Accept: text/markdown to any docs URL.
A map of every pagehttps://docs.fous.com/llms.txt
Every page in one filehttps://docs.fous.com/llms-full.txt
The API contracthttps://api.fous.com/openapi.json
Every public workflowhttps://fous.com/llms.txt, with a Markdown page per workflow

The Copy page button at the top of each page copies that page's Markdown. Its menu also opens the page in Claude or ChatGPT, or copies a prompt that points a coding agent at it.

Prompt a coding agent

Paste this into Claude Code, Codex, Cursor or any agent with web access. Change the page URL to the guide you are following.

Prompt
Read https://docs.fous.com/guides/coding-agents.md (the Fous docs page "Use with coding agents") and https://docs.fous.com/llms.txt. Then help me integrate Fous in this project, following those docs exactly. Ask me for anything the docs say I must decide, such as which workflow to call. Keep my Fous API key in an environment variable named FOUS_API_KEY.

For a whole integration, point the agent at llms.txt and let it fetch the pages it needs. An agent that has the MCP server connected can also search and run workflows while it writes the code, which is the fastest way to confirm an input schema.

Give the agent the MCP server

Claude Code, Codex, Cursor and VS Code all take the server in one line or one file. The install snippets, with one-click install links where the client supports them, are in Connect an AI agent.

What to tell the agent

  • Keep the key in FOUS_API_KEY, read it from the environment, and never commit it.
  • Send an Idempotency-Key with every POST /v1/query so retries never pay twice. Retries and idempotency.
  • Read success, then data.output; on failure branch on error.code, and respect error.retryable and Retry-After. Errors and limits.
  • Take inputs from the operation's input_schema in the catalog, never from its name. Find a workflow.