Introduction to MCP

Learn how to use Model Context Protocol to extend Cascade capabilities

Introduction to Model Context Protocol (MCP)

MCP (Model Context Protocol) empowers Windsurf's Cascade AI by connecting it to a wide array of third-party integrations and knowledge sources. This allows Cascade to access and utilize information from tools like Jira, Figma, GitHub, databases (e.g., Supabase), and more, directly within your workflow.

Why MCP?

  • Standardized Access: MCP standardizes how Cascade retrieves information from various local and remote data sources, eliminating the need for manual API calls or constant copy-pasting of context.
  • Enhanced Context: It brings crucial external context directly to Cascade, making its assistance more informed and powerful.
  • Windsurf as Client: Windsurf acts as the client, facilitating communication between Cascade and MCP servers.

Connecting to MCP Servers:

There are two main ways to connect Cascade to MCP servers:

  1. 1.

    Windsurf MCP Plugin Store:

    • Easy Connection: Discover and connect to verified MCP servers with a few clicks.
    • Example (Supabase):
      1. 1. Find "Supabase" in the plugin store and click "Install."
      2. 2. Paste your Supabase access token when prompted.
      3. 3. Once connected, refresh to see available tools (e.g., for reading/writing to your database).
    • Use Case: Cascade can then execute SQL queries via the Supabase MCP to perform tasks like designing database tables for an application.
  2. 2.

    Manual Configuration (via mcp_config.json):

    • For Unlisted Servers: If an MCP server isn't in the plugin store, you can configure it manually.
    • Process: Add the server's configuration details to your mcp_config.json file.
    • Example (Jira): Paste the Jira MCP server configuration into the file, and Cascade will gain access to Jira-specific tools (e.g., creating issues, searching tickets).
    • Finding Configurations: Search the web or MCP documentation for sample configurations for various servers.

Using MCP in Conversations

Automatic Detection: Cascade recognizes when MCP tools are relevant to your requests.

Explicit Usage: Ask Cascade to use specific MCP integrations: "Use the database MCP to query user data."

Tool Chaining: Cascade can combine multiple MCP tools in a single workflow.

Best Practices

  • Start with essential integrations your team uses daily
  • Test MCP plugins in development environments first
  • Review permissions and security settings regularly
  • Document custom MCP configurations for your team

Security Considerations

  • Use environment variables for sensitive credentials
  • Limit MCP plugin permissions to minimum required access
  • Monitor MCP usage and audit logs
  • Keep plugins updated to latest versions

MCP transforms Cascade from a code assistant into a comprehensive development orchestrator.