Browse documentation
All documentation

AI assistants (MCP)

Connect an AI assistant

Add the Fuzzbucket connector to Claude, ChatGPT, or another MCP-capable assistant, sign in, and choose what it can access.

Last reviewed October 15, 2026

Connecting takes about a minute. You copy one web address into your assistant, sign in to Fuzzbucket, and choose what the assistant is allowed to do. There is no password or code to copy for Claude, ChatGPT, and most other assistants.

Before you start

  • Have a Fuzzbucket account and know which email you use to sign in.
  • Use an assistant that supports custom MCP connectors. Claude, ChatGPT, Claude Code, Cursor, VS Code, Gemini CLI, and Grok all do. Some assistants only offer custom connectors on certain plans.
  • Keep the Fuzzbucket server URL ready: https://api.fuzzbucket.ai/mcp

You can also copy it from the Fuzzbucket MCP page or from Settings > MCP inside Fuzzbucket.

Step 1: Add Fuzzbucket to your assistant

Pick your assistant below. The Fuzzbucket MCP page shows the same steps for every supported assistant, with a copy button for the address or command.

Claude (web, desktop, and mobile)

  1. In Claude, open the connector settings and choose Add custom connector. Depending on your version, connectors are under Settings or Customize.
  2. Name it Fuzzbucket and paste the server URL.
  3. Select Connect. A Fuzzbucket sign-in window opens.

ChatGPT (web and apps)

  1. Open the area where ChatGPT manages apps or plugins and choose to add a custom MCP server. Its name depends on your version and plan.
  2. Paste the server URL and choose OAuth sign-in.
  3. Create and install the plugin. A Fuzzbucket sign-in window opens.

Other assistants

Add the server URL as a remote HTTP MCP server, then sign in when the assistant asks. Coding agents such as Claude Code, Codex, Cursor, VS Code, and Gemini CLI have two options: run the one-line command or install button on the Fuzzbucket MCP page and sign in, or paste a ready-made setup prompt from Settings > MCP that already includes a token. See Manage connected assistants and tokens.

Step 2: Sign in and choose access

After you connect, Fuzzbucket asks you to approve the assistant. Check that the signed-in email is the account you want, then review the screen before you continue.

  • What it can view lists the information the assistant can read.
  • What it can do has three levels. Read only cannot change anything. Create and edit can create workspaces and edit Canvas drafts but cannot generate. Full access can also chat with the orchestrator and generate content, which spends your credits.
  • Brand access is set to All brands by default. Choose Only selected brands to keep the assistant inside the brands you pick.

Select Allow access. You return to your assistant, and the connector is ready.

Only approve apps you trust. If Fuzzbucket warns that an app is unverified or that the request was started somewhere else, cancel unless you started it yourself.

Step 3: Try your first request

Start a new conversation in your assistant and ask something simple, such as "Which brands do I have in Fuzzbucket?" If it answers with your brands, everything works.

Then try a real task, for example "Draft a product launch post for my brand in the spring campaign workspace." Open the workspace in Fuzzbucket to see the drafts and results the assistant created.

If something does not work, see the troubleshooting section in Manage connected assistants and tokens.