Set up Maze MCP

Claude (Desktop App and claude.ai)

For Pro and Max plans:

1. Go to your Claude account

2. Click on Customize in the left sidebar and then Connectors

3. Click on “Browse” and search for Maze and select it

4. Click Connect

For Team and Enterprise plans:

Preliminary steps for owners: Before members of Team and Enterprise plans can configure custom connectors, an Owner or Primary Owner needs to follow these initial steps to add a custom connector to your organization:

1. Go to your Claude account and navigate to Organization settings > Connectors

2. Click on “Browse” and search for Maze and select it

3. Click Connect

Steps for members after connector is configured:

1. Go to your Claude account

2. Click on Customize in the left sidebar and then Connectors

3. Click on “Browse” and search for Maze and select it

4. Click Connect

ChatGPT

1. Go to your ChatGPT account

2. Click on Apps in the left sidebar

3. Search for Maze and select it

4. Click Connect

Cursor

1. Open Cursor and go to Settings → Tools & MCPs

2. Click Add Custom MCP

3. Paste the following config under mcpServers:

{
  "mcpServers": {
    "maze": {
      "url": "https://connect.maze.co/mcp"
    }
  }
}

4. Save, then return to Tools & MCPs and click Connect on the Maze entry

Gemini CLI

1. Install Gemini CLI (needs Node.js 20+)
   npm install -g @google/gemini-cli

2. Add the Maze server to ~/.gemini/settings.json (mac) or C:\Users\<username>\.gemini\settings.json (windows)

{
  "mcpServers": {
    "maze": {
       "httpUrl": "https://connect.maze.co/mcp"
    }
  }
}

1. Authenticate. Launch it by running gemini in your terminal and then in the CLI run: /mcp auth mazeA

2. The browser window will open. Sign in with your Maze account. The token is stored and reconnects automatically after that.

3. Verify. Run /mcp to confirm the Maze server shows as connected with its tools listed.

GitHub Copilot (VS Code)

Option A — Command Palette:

1. Open the Command Palette (⇧⌘P on Mac / Ctrl+Shift+P on Windows/Linux) and run MCP: Add Server

2. Choose HTTP (Streamable)

3. Enter the server URL: https://connect.maze.co/mcp

4. Name it maze-connect and choose Global (all workspaces) or Workspace

Option B — Edit config directly:

Run MCP: Open User Configuration for a global setup, or create .vscode/mcp.json in your workspace, and add:

{
  "servers":  {
    "maze": {
      "type": "http",
      "url": "https://connect.maze.co/mcp"
    }
  }
}

Then:

5. Click Start on the CodeLens above the server entry in mcp.json

6. Click Auth to complete the OAuth flow in your browser

7. In the Chat view, switch the mode picker to Agent — Maze tools appear under Configure Tools in the chat input (or type # to reference them)

Verify it works: Run "Which Maze teams am I in?" — Copilot should respond with your team list.

Microsoft Copilot

To enable Maze MCP inside Copilot and other Microsoft channels (like Microsoft Teams) you need to create a new agent with a custom tool connector that points to our MCP:

1. Go to https://copilotstudio.microsoft.com/ and sign in

2. Go to the Agents tab and Click on Create blank agent

3. Enter the name of the agent; we recommend Maze

4. On the overview page you can setup a user facing description, choose the model you want to use for the agent and enter custom instructions to further customize the agent. We recommend also disabling web search since Maze MCP doesn’t need it

5. Go to the Tools section and click Add a tool

6. Select the Model Context Protocol card and enter:

   1. Server name - Maze Connect

   2. Server description - Connect Maze to look up research studies, review session results, and surface user insights without leaving the workflow. Find studies by topic even when the study name is unknown, go through participant responses and task metrics, retrieve highlights and themes from moderated and unmoderated sessions, and pull recordings or transcripts for deeper analysis. This is useful for pulling evidence for a product decision, synthesizing research before a design review, or catching up on studies that have not been reviewed yet.

   3. Server URL - https://connect.maze.co/mcp

   4. Authentication - Select OAuth 2.0 and type Dynamic discovery

   5. Click create

   6. Once created click on the Connection and create a new connection - you will be prompted for your Maze Account and authorizing access for the MCP

   7. Once the connection is set click Add and configure

7. If everything set up correctly, you should see the tool and resources list populated

   1. If the agent was still provisioning you might need to refresh the tool list and resources list

8. Use the right chat box to test your agent (try: “What maze teams am I apart of” or some other prompt from the help article)

   1. You will be asked to Open connection manager to setup the connection for the user. This is because the Agent uses the users Maze permissions instead of a service level account. Once done hit retry and check the MCP returns a result

9. On the channels tab choose which channel you want to provision your agent for

10. Hit publish to publish your agent

    1. Depending on your organization settings, you might need an admin to approve the agent to appear in the correct channels

Other AI apps

Claude Code (CLI)

1. Run the following in your terminal:

claude mcp add --transport http maze-connect https://connect.maze.co/mcp

2. Open a Claude session

3. Run /mcp and select maze-connect (it will indicate authentication is needed)

4. Click Authenticate and follow the auth flow in your browser

Codex

1. Open the Codex app and run:

codex mcp add maze-connect --url https://connect.maze.co/mcp

2. If not automatically prompted to log in, run:

codex mcp login maze-connect

3. Fully close the Codex app (don't just minimize it) and reopen it — Maze MCP will now be available

Learn more

• Maze MCP Overview

• Maze MCP Data & Privacy

Still need help?

If you have any questions or concerns, please let our Support team know — we'll be happy to help!