Firestore MCP Server

What is Firestore MCP Server?

An MCP (Model Context Protocol) server for interacting with Google Firestore directly through Claude Desktop. It provides a clean interface for creating, reading, updating, and deleting Firestore documents, targeting developers who use Claude and need to manage Firestore data programmatically.

How to use Firestore MCP Server?

Install dependencies with npm install, build with npm run build, then configure Claude Desktop by adding the server to claude_desktop_config.json with the path to build/index.js, a GOOGLE_CLOUD_PROJECTS environment variable listing project IDs, and corresponding .json credential files in a keys folder. The service account must have Firestore permissions (e.g., Cloud Datastore Owner). Tools are invoked via natural language prompts in Claude Desktop (e.g., "Get the document with ID user123 from the users collection").

Key features of Firestore MCP Server

• Create documents in Firestore collections
• Read documents by ID from a collection
• Update existing documents
• Delete documents from a collection
• Query documents with filtering, ordering, and limits
• List all available Firestore collections

Use cases of Firestore MCP Server

• Retrieve a user document by ID for profile display
• Create a new user record in a Firestore collection
• Update a document field (e.g., change age) through natural language
• Delete a document when a user account is removed
• Query and list users matching certain criteria (e.g., over 25, ordered by name)

FAQ from Firestore MCP Server

What dependencies does Firestore MCP Server require?

Node.js and npm are required to install and build the project. You also need a Google Cloud project with Firestore enabled and a service account with appropriate permissions.

How does authentication work?

Authentication uses service account JSON credential files placed in a keys folder, one per project. The server reads these files to authenticate with Google Firestore. The GOOGLE_CLOUD_PROJECTS environment variable lists the project IDs.

Where does Firestore data live?

All data resides in your Google Cloud Firestore database managed by the listed projects. The server only reads and writes to Firestore; no local storage is used.

What transports or protocols does the server use?

The server communicates with Claude Desktop via the Model Context Protocol (MCP). No additional transports (HTTP, WebSocket, etc.) are mentioned.

Are there any known limits or restrictions?

The README does not specify limits. However, the server depends on Google Cloud Firestore quotas and permissions. The first project in GOOGLE_CLOUD_PROJECTS is treated as the default.