Skip to main content

​
Getting Started with Goose Desktop

Configure Collate’s MCP Server to interact with block’s AI agent Goose.

​
Prerequisites

For this guide, you will need:

​
Adding your Collate MCP Server to Goose Desktop

This how-to guide uses Goose Desktop for macOS. Make sure that you already have an LLM Provider configured before prompting.
  • Navigate to Goose Desktop’s Settings, then under Extensions, select +Add custom extension.
Goose settings
  • The custom extension should have the following information:
    • Extension Name Collate
    • Command paste the following command: 
      npx -y mcp-remote <YOUR_Collate_SERVER>/mcp --auth-server-url=<YOUR_Collate_SERVER>/mcp --client-id=Collate --verbose --clean --header Authorization:${AUTH_HEADER}
      • If you are running it locally, your command will look like this: 
        npx -y mcp-remote http://localhost:8585/mcp --auth-server-url=http://localhost:8585/mcp --client-id=Collate --verbose --clean --header Authorization:${AUTH_HEADER}
    • Add 1 Environment Variable
      • Variable name is AUTH_HEADER
      • Value is Bearer &lt;YOUR_Collate_PAT&gt;
Configuring Collate MCP Server
  • Select +Add to store this Environment Variable
Configuring Collate MCP Server
  • Select Add Extension
Adding extension Collate successfully added

​
Prompt to read from Collate

This part of the guide assumes that you have assets in Collate. You can add data assets into Collate here. Select a model from Goose and paste the following prompt to have it read from Collate: 
What tables do you have access to in Collate?
For more sample use cases with MCP please check out our blog!

​
Reach out on Slack!

With MCP, we are finding new ways to use Collate all the time! Now that you have Claude and Collate configured to work together, think you’ve got a great new use case? Show us what you’ve got in Slack!