Connect your agent to Couldi Manage.

Manage works with any agent that speaks MCP. Pick the app you use and follow its steps. You approve every connection on a Couldi page, and you choose which boards each agent can reach.

Before you start

One address, one approval.

Every app below connects to the same address. You need access to Couldi Manage first:get access on the Manage page.

https://mcp.couldi.xyz/mcp

Jump to: Claude app · Cursor · Claude Code · Codex · ChatGPT · VS Code · Gemini CLI

claude.ai or Claude Desktop. No terminal and no config file.

Claude app

  1. In Claude, open Customize → Connectors, press + and choose Add custom connector.

  2. Paste this address as the server URL, then press Add:

    https://mcp.couldi.xyz/mcp
  3. When Claude asks you to connect, allow it. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

Claude's free plan allows one custom connector. On a Team or Enterprise plan, an Owner adds it first under Organization settings → Connectors.

Claude app's own instructions

One click, with Cursor installed on this computer.

Cursor

  1. Press Add to Cursor and confirm the install when Cursor asks.

    Add to Cursor
  2. When Cursor asks you to sign in to couldi, approve it. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

Cursor's own instructions

In a terminal.

Claude Code

  1. In the folder your project lives in, run:

    claude mcp add --transport http couldi https://mcp.couldi.xyz/mcp
  2. Start Claude Code, type /mcp, choose couldi and sign in. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

Claude Code's own instructions

The Codex app, its IDE extension, or the Codex CLI.

Codex

  1. In the Codex app or IDE extension, open MCP servers and choose Add server. Name it couldi, pick Streamable HTTP and paste this address:

    https://mcp.couldi.xyz/mcp
  2. Save and restart Codex, then press Authenticate next to couldi. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

  3. Using the Codex CLI instead? Run these two lines:

    codex mcp add couldi --url https://mcp.couldi.xyz/mcp
    codex mcp login couldi

Codex's own instructions

Developer mode, on the web.

ChatGPT

  1. In ChatGPT on the web, open Settings → Security and login and turn on Developer mode.

  2. Open ChatGPT's Plugins page and press + to create an app.

    Open ChatGPT Plugins
  3. Name it Couldi, paste this address as the MCP server URL, choose OAuth and create it:

    https://mcp.couldi.xyz/mcp
  4. A Couldi page opens: sign in, choose the boards the agent may reach, and approve. Then, in a chat, choose Developer mode from the + menu and select Couldi.

Developer mode is on ChatGPT's Plus, Pro, Business, Enterprise and Education plans, on the web only. On a Business or Enterprise workspace, an admin has to allow it first.

ChatGPT's own instructions

GitHub Copilot agent mode, with the code command on your path.

VS Code

  1. Run:

    code --add-mcp '{"name":"couldi","type":"http","url":"https://mcp.couldi.xyz/mcp"}'
  2. When VS Code asks to sign in to couldi, allow it. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

VS Code's own instructions

In a terminal.

Gemini CLI

  1. Add this to ~/.gemini/settings.json, merging it with any mcpServers you already have:

    {
      "mcpServers": {
        "couldi": {
          "httpUrl": "https://mcp.couldi.xyz/mcp"
        }
      }
    }
  2. Start Gemini CLI and run /mcp auth couldi. A Couldi page opens: sign in, choose the boards the agent may reach, and approve.

Gemini CLI's own instructions

After you connect

What the agent can do, and how to stop it.

The approval page lists what the agent will be able to do: read your boards, propose cards, and, if it asks, move and reject them. A proposal lands in Maybe for you to rule on, and any move an agent makes is flagged for your review. Disconnect an agent at any time from Manage → Connected agents.

Manage stores what agents write to your boards, and never sees your agent's own prompts. The details are in the privacy policy. Questions: support@asc.is.