kitty over MCP

kitty runs a remote Model Context Protocol server, so an AI assistant can work your ventures directly: read a verdict, pull a board's research, ask an advisor, add todos. Anything the assistant does is done as you — same workspace, same per-venture access, same plan limits and Treats as the web app.

1. Create an API key

Go to Settings → Security → API keys and create one. The key is shown once — copy it then. It starts with kitty_sk_, it does not expire, and you can revoke it from the same place at any time.

Treat it like a password: it can read and change everything you can. It cannot create more keys.

2. Point your client at it

The endpoint is:

https://kitty.build/api/mcp

Claude Code

claude mcp add --transport http kitty https://kitty.build/api/mcp \
  --header "Authorization: Bearer kitty_sk_YOUR_KEY"

Claude Desktop, Cursor and other clients that read a config file

{
  "mcpServers": {
    "kitty": {
      "type": "http",
      "url": "https://kitty.build/api/mcp",
      "headers": {
        "Authorization": "Bearer kitty_sk_YOUR_KEY"
      }
    }
  }
}

Claude API

{
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://kitty.build/api/mcp",
      "name": "kitty",
      "authorization_token": "kitty_sk_YOUR_KEY"
    }
  ]
}

What your assistant can do

• list_ventures, get_venture, get_portfolio — see what you have and what needs attention
• get_board_report — the full written research from any of the nine boards
• create_venture, start_analysis, get_analysis_status — submit an idea and follow the run
• ask_advisor — the advisors, with your venture's board analysis in context
• manage_todos, add_note — capture and close work
• get_credit_balance — Treats remaining and what each action costs

Good to know

Deep analysis takes minutes. start_analysis returns immediately and the nine boards run in the background — your assistant should poll get_analysis_status rather than wait.

It costs the same as the app. A deep analysis over MCP spends the same Treats as one started in the browser. If you run out, the tool says so and links you to top up.

Requests are rate limited to 60 per minute per key — plenty for real work, and enough to stop a looping agent.

Trouble? Email info@foundergem.com with the client you are using and we'll take a look.