Firebase MCP server

The Firebase MCP server is a local Model Context Protocol (MCP) server that runs in your development environment to give AI-powered development tools the ability to work with your Firebase projects and your app's codebase.

The Firebase MCP server works with any tool that can act as an MCP client, including: Antigravity, Claude Code and Claude Desktop, Cline, Cursor, VS Code Copilot, Windsurf, and more!

arrow_downward Jump to setup instructions

Benefits of the MCP server

An editor configured to use the Firebase MCP server can use its AI capabilities to help you:

• Create and manage Firebase projects
• Manage your Firebase Authentication users
• Work with data in Cloud Firestore and Firebase SQL Connect
• Retrieve Firebase SQL Connect schemas
• Understand your security rules for Cloud Firestore and Cloud Storage for Firebase
• Send messages with Firebase Cloud Messaging

These are only partial lists; see the server capabilities section for a complete list of tools available to your editor.

Set up your MCP client

The Firebase MCP server is a local MCP server that can work with any MCP client that supports standard I/O (stdio) as the transport medium.

When the Firebase MCP server makes tool calls, it runs locally in your environment using the same user credentials that authorize the Firebase CLI. This could be a logged-in user or Application Default Credentials, depending on the environment.

Before you begin

Make sure you have a working installation of Node.js and npm.

Basic configuration

Here are basic configuration instructions for using the Firebase MCP server with some popular AI-assistive tools:

{
  "mcpServers": {
    "firebase-mcp-server": {
      "command": "npx",
      "args": ["-y", "firebase-tools@latest", "mcp"]
    }
  }
}

{
  "mcpServers": {
    "firebase": {
      "command": "npx",
      "args": ["-y", "firebase-tools@latest", "mcp"]
    }
  }
}

"mcpServers": {
  "firebase": {
    "command": "npx",
    "args": ["-y", "firebase-tools@latest", "mcp"]
  }
}

{
  "mcpServers": {
    "firebase": {
      "command": "npx",
      "args": ["-y", "firebase-tools@latest", "mcp"],
      "disabled": false
    }
  }
}

"servers": {
  "firebase": {
    "type": "stdio",
    "command": "npx",
    "args": ["-y", "firebase-tools@latest", "mcp"]
  }
}

"mcp": {
  "servers": {
    "firebase": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "firebase-tools@latest", "mcp"]
    }
  }
}

"mcpServers": {
  "firebase": {
    "command": "npx",
    "args": ["-y", "firebase-tools@latest", "mcp"]
  }
}

Optional configuration

In addition to the basic configuration for each client, shown earlier, there are two optional parameters you can specify:

• --dir ABSOLUTE_DIR_PATH: The absolute path of a directory containing firebase.json, to set a project context for the MCP server. If unspecified, the get_project_directory and set_project_directory tools become available and the default directory will be the working directory where the MCP server was started.

• --only FEATURE_1,FEATURE_2: A comma-separated list of feature groups to activate. Use this to limit the tools exposed to only features you are actively using. Note that the core tools are always available.

For example:

"firebase": {
  "command": "npx",
  "args": [
    "-y",
    "firebase-tools@latest", "mcp",
    "--dir", "/Users/turing/my-project",
    "--only", "auth,firestore,storage"
  ]
}

MCP server capabilities

The Firebase MCP server provides three different categories of MCP features:

• Prompts: A library of pre-written prompts that you can run; they're optimized for developing and running an app with Firebase

• Tools: A set of tools intended for use by AI agents that help them work directly with your Firebase project (with your approval!)

• Resources: Documentation files intended for use by AI agents to give them more guidance and context to complete a task or goal

Prompts

The Firebase MCP server comes with a library of pre-written prompts optimized for developing and running an app with Firebase. You can use these prompts to complete various common tasks or goals with your agentic AI assistants.

The following table describes the prompts the MCP server makes available.

Many development tools that support MCP provide a convenient way to run these prompts. For example, as a slash command:

/firebase:init

In supported environments, you can start typing /firebase: to see a list of available prompts.

npx firebase-tools@latest mcp --generate-prompt-list

Tools

The Firebase MCP server also provides a number of tools intended for use by AI agents that help them work directly with your Firebase project (with your approval!). Unlike with prompts, you don't call these tools directly; rather, AI agents that support tool calling (such as Antigravity, Claude Code, or Cursor) can automatically call these tools to perform development tasks when needed.

The following table describes the tools the MCP server makes available.

npx firebase-tools@latest mcp --generate-tool-list

{
  "parent": "projects/my-project/databases/(default)/documents",
  "structured_aggregation_query": {
    "structured_query": {
      "from": [{"collection_id": "orders"}],
      "where": {
        "field_filter": {
          "field": {"field_path": "status"},
          "op": "EQUAL",
          "value": {"string_value": "COMPLETED"}
        }
      }
    },
    "aggregations": [
      {
        "alias": "total_orders",
        "count": {}
      },
      {
        "alias": "average_amount",
        "avg": {
          "field": {"field_path": "amount"}
        }
      }
    ]
  }
}

Resources

The MCP server provides resources, which are documentation files intended for use by AI agents. Agents that support using resources will automatically include relevant resources in the session context.

The following table describes the resources the MCP server makes available.

npx firebase-tools@latest mcp --generate-resource-list