This feature is currently only available for paying individual users. It is currently not available for Free, Teams, or Enterprise users.

MCP (Model Context Protocol) is a protocol that enables LLMs to access custom tools and services. Put briefly, an MCP client (Cascade, in this case) can choose to make requests to MCP servers to access tools that they provide. Cascade now natively integrates with MCP, allowing you to bring your own selection of MCP servers for Cascade to use. See the official MCP docs for more information.

Configuring MCP with Cascade

To set up MCP with Cascade, you will need to specify a list of MCP servers along with any necessary arguments and environment variables that Cascade will need to run the servers. This list should be specified as a JSON file at ~/.codeium/windsurf/mcp_config.json. The JSON should follow the same schema as the config file for Claude Desktop.

For quick access to this file, you can find the toolbar above the Cascade input, click the hammer icon, and then click the “Configure” button that appears.

A view of the MCP toolbar with no servers configured.

Clicking “Configure” will open the mcp_config.json file in the IDE. Here’s an example configuration, which sets up a single server for Google Maps:

{
  "mcpServers": {
    "google-maps": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-google-maps"
      ],
      "env": {
        "GOOGLE_MAPS_API_KEY": "<YOUR_API_KEY>"
      }
    }
  }
}

Be sure to provide the required arguments and environment variables for the servers that you want to use.

Once you have configured the JSON file, click “Refresh” in the toolbar. If the integration is successful, the toolbar should indicate that you have MCP servers available.

Click the line with the server name to see an expanded view that lists the server's tools.

See the official MCP server reference repository or OpenTools for some example servers.

Notes

  • Since MCP tool calls can invoke code written by arbitrary server implementers, we do not assume liability for MCP tool call failures. To reiterate:

MCP tool calls will consume credits regardless of success or failure.

  • We currently only support tools, not prompts nor resources. In other words, Cascade will be able to request and use a server’s tools, but not the other endpoints that a server exposes.
  • We do not support tools that output images.
  • We only support servers using the stdio transport type.