Firebase Studio is a powerful, agentic development environment, designed to streamline your workflow with AI assistance. It provides code suggestions, generates code, explains concepts, updates multiple files, and even runs terminal commands. But what if you could extend these AI capabilities even further by integrating external tools and data sources? This is where Model Context Protocol (MCP) servers come in.

What’s MCP?

The Model Context Protocol (MCP) provides a standardized way for AI-native development environments like Firebase Studio (MCP Host/Client) to connect to a wide range of external tools and data sources (MCP servers). MCP servers provide additional tools and data sources for AI to use, enhancing its ability to interact with your workspace. This means Gemini in Firebase Studio can become even more knowledgeable and capable, leveraging these external resources to offer more insightful assistance and automate complex tasks.

Firebase Studio’s initial support for MCP servers is focused on empowering power-users to supercharge their AI-assisted development workflow with external tool integration and data sources.

MCP server configuration

To add an MCP server in FIrebase Studio, create a .idx/mcp.json file in your project and add the MCP server’s configuration to the file.

Here’s an example of a MCP server configuration:

{
  "mcpServers": {
    "server-name": {
      "command": "npx",
      "args": ["-y", "mcp-server"],
      "env": {
        "key-name": "value"
      }
    }
  }
}

For more information about MCP server configuration, see Customize your Firebase Studio workspace.

Prefer hands-on learning? Complete the Firebase MCP in Firebase Studio codelab.

Supported MCP server types and limitations

Firebase Studio currently supports MCP servers that use standard input/output (stdio) or Server-Sent Events (SSE)/Streamable HTTP transport that don’t require special forms of authentication. You can also pass environment variables to MCP servers through the .env file.

However, Firebase Studio currently does not support MCP servers that require a graphical user interface or a desktop session. This means servers controlling full browser instances, for example, will not function in the constrained Firebase Studio VM. The current implementation focuses solely on the “tools” capability of the MCP protocol.

The future of MCP in Firebase Studio

This foundational release is just the beginning, and we’re incredibly excited to see what you build with it1 The future of AI extensibility in Firebase Studio will be shaped by your real-world use cases and creative feedback. Join the conversation in our community forum to share what you’re making and inspire what comes next!