Firebase MCP enables AI assistants to work directly with Firebase services, including Firestore, Storage, and Authentication. It exposes MCP-compatible tools for common Firebase operations: Firestore (add/list/get/update/delete documents, list root collections, query across subcollections), Storage (list files, get file info, upload content, upload from URL), and Authentication (get user by identifier). The server supports both stdio and HTTP transports, allowing deployment as a standalone HTTP service with session management so multiple clients can share a single server instance. Configuration relies on a Firebase service account key and an optional storage bucket, supplied via environment variables such as SERVICE_ACCOUNT_KEY_PATH and FIREBASE_STORAGE_BUCKET. The MCP server can be run via npm, using npx or a local installation, with guidance for configuring transport and port settings (MCP_TRANSPORT, MCP_HTTP_PORT, MCP_HTTP_HOST, MCP_HTTP_PATH). It also provides developer tooling and testing guidance, including Firebase emulator usage, test commands, and an MCP Inspector for interactive debugging. Client integrations include Claude Desktop, Augment Code, VS Code, and Cursor. The project includes troubleshooting, logging, and a comprehensive API reference for all supported tools.