Connect your AI assistant to Cabi.Chat (MCP)

Your AI assistant can work in Cabi.Chat for you. Cabi.Chat runs an MCP server: Claude Code, Codex, Cursor, Claude Desktop or any other MCP client can find Cabi.Chat API operations, read how they work and run them on your account. Tell the assistant what you want in plain words and it makes the calls. What it can do for you: Let your AI agent run your presence on Cabi.Chat

Before you start

You need a personal access token (an API key): Cabi.Chat API — Personal access tokens. Mint one: API pipeline: Personal access tokens

The server

• Address: https://api.cabi.chat/mcp
• Transport: Streamable HTTP
• Sign-in: your token in every request, as the header Authorization: Bearer YOUR_TOKEN

Add it to your assistant

Claude Code

claude mcp add --transport http cabi-chat https://api.cabi.chat/mcp --header "Authorization: Bearer YOUR_TOKEN"

Run /mcp in Claude Code to check that cabi-chat is connected.

Codex

Put the token in an environment variable, for example CABI_TOKEN, and add this to ~/.codex/config.toml:

[mcp_servers.cabi-chat]
url = "https://api.cabi.chat/mcp"
bearer_token_env_var = "CABI_TOKEN"

Cursor

Add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "cabi-chat": {
      "url": "https://api.cabi.chat/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Claude Desktop

Claude Desktop reaches a server that takes a token header through the mcp-remote bridge (needs Node.js). Add this to claude_desktop_config.json and restart Claude Desktop:

{
  "mcpServers": {
    "cabi-chat": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.cabi.chat/mcp", "--header", "Authorization:${CABI_AUTH}"],
      "env": { "CABI_AUTH": "Bearer YOUR_TOKEN" }
    }
  }
}

The server's tools

The server offers three tools:

• search_api_endpoints - finds Cabi.Chat API operations for a task described in plain words, best match first.
• get_api_endpoint - reads one operation in full: method, path, parameters, request body and response.
• call_api - runs one operation and returns the status and the response.

A typical run: the assistant searches for what you asked, reads the operation, then calls it.

The assistant acts as you

What it creates belongs to you, and anything paid is charged to your balance, the same as when you do it yourself. Read what your assistant plans to do before you let it make changes. What a token can do: Cabi.Chat API — Personal access tokens

If something does not work

• The assistant reports that sign-in is required: the token is missing or no longer valid. Check the Authorization: Bearer YOUR_TOKEN header, or create a new token.
• An operation is not available: the assistant asked for something this server does not offer to your account. Ask it to search again for your task.

More about the API itself: Cabi.Chat API