Anahita MCP

Anahita's docs and API reference, in the AI agent you already use: Claude, Gemini, Cursor, VS Code, or any other that takes an MCP server. The agent then answers from the docs as they are today, not from what it remembers: how to set up a site, how federation works, which route follows a profile and what it answers.

The server is anahita-mcp (opens in a new tab), open source under MIT. It reads, and only reads: these docs, the web app's, and the API reference. There are two ways to use it.

Way	Address or command	For
Hosted, nothing to install	https://mcp.anahita.io/mcp	Chat apps, which reach servers over the internet, and anybody without Node.js
On your computer	npx -y anahita-mcp	Coding agents; your questions never leave your machine through it. Needs Node.js (opens in a new tab) 20 or later

Adding it

Client	Hosted	On your computer
Claude Code	claude mcp add --transport http anahita https://mcp.anahita.io/mcp	claude mcp add anahita -- npx -y anahita-mcp
Claude Desktop and claude.ai	Settings › Connectors › Add custom connector, with the address	Settings › Developer › Edit Config: in mcpServers, "anahita": { "command": "npx", "args": ["-y", "anahita-mcp"] }, then restart
Gemini CLI	gemini mcp add --scope user --transport http anahita https://mcp.anahita.io/mcp	gemini mcp add --scope user anahita npx -- -y anahita-mcp: the -- keeps -y for npx, as the Gemini CLI has a -y of its own
Cursor	In ~/.cursor/mcp.json, under mcpServers: "anahita": { "url": "https://mcp.anahita.io/mcp" }	"anahita": { "command": "npx", "args": ["-y", "anahita-mcp"] }
VS Code	In .vscode/mcp.json, under servers: "anahita": { "type": "http", "url": "https://mcp.anahita.io/mcp" }	"anahita": { "type": "stdio", "command": "npx", "args": ["-y", "anahita-mcp"] }
Any other	The address, as a remote MCP server (Streamable HTTP)	The command, as a local one (standard input and output)

Then ask, for example: "How does Anahita decide what an old post is worth?", or "Which API route lets a person follow a group, and what does it send back?"

What it can do

Tool	Answers
search_docs	The sections of the docs that match some words, best first, with where each is
list_docs	Every page, in its groups, with its sections
read_doc	A page, or one section of it
list_endpoints	The API's services; one service's routes; or the routes that match some words
read_endpoint	One route in full: who may call it, what it takes, every answer it gives
read_shape	One shape the API takes or answers with, and its fields

Every answer gives the www.anahita.io (opens in a new tab) address of what it read, so that the agent can point you to the page.

What it keeps

• On your computer: nothing is sent anywhere. It fetches two files from www.anahita.io (opens in a new tab), keeps them in your system's cache folder, and searches them there. It works offline with what it fetched last.
• Hosted: it keeps no record of the questions or of who asked. AWS keeps only when the function ran and for how long, for one day. It runs in Montreal (AWS ca-central-1).

The docs it reads are a day behind these at most: the website is rebuilt on every change to it and every night.

Without MCP

File	What it holds
www.anahita.io/llms.txt (opens in a new tab)	The site and its docs in brief, for language models
www.anahita.io/docs/index.json (opens in a new tab)	Every page of the docs as Markdown, cut into sections, in one file
www.anahita.io/api/anahita.json (opens in a new tab)	The API reference, OpenAPI 3.1

Your own copy

An operator can host the server for their own site's docs, on any server (npx -y anahita-mcp --http) or on AWS Lambda: Hosting it (opens in a new tab).

Where the code is

What	Where
The server	anahita-mcp (opens in a new tab): src/server.ts for the tools
The docs index it reads	anahita.io: src/app/docs/index.json/route.js
The API reference it reads	docs/api/, made from the code (The API)