Firebase MCP Server

Official

Firebase-specific capabilities for AI coding tools

How to install ↓ Compare GitHub source repository ↗

Category: Dev Tools #133 of 438
Stars: ★ 4.5k Very popular
Transport: stdio (local process)
Runtime: Node.js
Credentials: API key / credential required
License: MIT
Last commit: today

59FMRS · C

This server is shipped directly from the official Firebase repository (firebase/firebase-tools), running the firebase-tools npm package's mcp subcommand over stdio. The source material does not list individual tool names or confirm support for clients beyond the Cursor one-click install shown in the README. Users should verify Firebase authentication is configured and be aware the underlying CLI can perform high-privilege actions like deployments and data deletion.

Strongest · Maintenance 16/20 Weakest · Security and permissions 9/20

Reliability

10/20

Security and permissions

9/20

Maintenance

16/20

Documentation

11/20

Setup experience

13/20

Why each score

Reliability 10/20: The manifest declares a coherent npm package ([email protected]), an npx runtime hint, a fixed positional argument "mcp", and stdio transport, which is internally consistent and matches the README's mention of a 'Firebase MCP Server'. However, the supplied material contains no verifiable execution evidence — no CI workflow content, no unit/integration tests, and no MCP tool listing to cross-check declared behavior against the description, so the init/tool-list handshake cannot be confirmed to work in practice. Per the static calibration rule, reliability is capped at 12 without such evidence, so this scores 10.
Security and permissions 9/20: The underlying Firebase CLI exposes many high-privilege, irreversible operations (deploy, database:remove, recursive firestore:delete, hosting:disable, functions:secrets:destroy, auth:import overwriting user data) and supports long-lived FIREBASE_TOKEN credentials alongside service-account keys and ADC. The README documents credential precedence reasonably transparently, but the provided material says nothing about which tools the MCP server actually exposes, whether destructive operations require confirmation, or how scope/least-privilege is enforced — a real undisclosed risk gap. No evidence of malware, fabricated identity, or leaked secrets was found, so this is not a red-line 0-4 case, but the unresolved confirmation/scoping gap keeps the score at 9.
Maintenance 16/20: This is the official Firebase CLI repository, with 4450 stars, 990 open issues, not archived, MIT-licensed, a visible GitHub Actions CI badge in the README, and an independently versioned mcp subfolder (0.3.0) alongside a mature main package version (14.27.0) — all indicating active, ongoing maintenance. No dedicated security-response channel is described in the given material, and the open-issue count is sizable, so it falls short of a perfect score, landing at 16.
Documentation 11/20: The README thoroughly documents general Firebase CLI commands, authentication modes, CI usage, proxy configuration, and programmatic use, but MCP-specific documentation is reduced to a single pointer link to the ./src/mcp directory, with no tool parameters, limits, cost, or troubleshooting content included in what was provided. This documentation chain is broken at the point that matters most for this review, so it scores 11.
Setup experience 13/20: The manifest specifies a clear npx-based launch with a fixed argument, and the README offers npm global install, a standalone binary installer, and a one-click Cursor install deep link, giving a reasonably complete, low-step setup path. However, there is no evidence of an actual successful 'npx firebase-tools mcp' handshake in any environment, and no generic JSON client-config example is shown beyond the Cursor button. Per the static calibration rule, setup is capped at 15 absent verifiable execution evidence, so this scores 13.

Static review · not runListed 2026-08-06

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the networkChanges third-party account dataConnects to a database

Best for

Developers already using the firebase-tools CLI who want an AI assistant to perform common Firebase operations
Teams doing Firebase integration work from within a local project directory

Not for

Users without an existing Firebase project or local authentication set up
Teams that require strict approval gates before an AI agent can trigger production deploys or data deletions

Required permissions

Local filesystem access (to read the project directory and files like firebase.json)
Firebase/Google Cloud credentials (user login session, service account key, or application default credentials)
Network access to the target Firebase project

Risks and side effects

firebase-tools supports destructive operations (e.g. firestore:delete, database:remove) and production deploys, so overly broad credentials could let an AI agent cause data loss or affect a live project
Service account keys and long-lived user tokens are highly sensitive credentials if leaked
Neither the README nor server.json enumerates the specific MCP tools exposed, so the exact operation surface should be verified before use

Setup

Before you start

Runtime：Node.js

GOOGLE_APPLICATION_CREDENTIALS optionalsecret Path to a JSON service account key file used for authentication; create the key via the Google Cloud Console.

FIREBASE_TOKEN optionalsecret Long-lived user token generated by firebase login:ci (deprecated; prefer service accounts), used in non-interactive environments.

Other optional settings (2)

HTTPS_PROXY optional URL of your HTTPS proxy (e.g. http://127.0.0.1:12345), set when accessing Firebase behind a proxy.

HTTP_PROXY optional URL of your HTTP proxy, set when accessing Firebase behind a proxy.

Install Node.js/npm locally.
Authenticate with firebase login, or set GOOGLE_APPLICATION_CREDENTIALS to a service account key file.
Add a stdio MCP server entry pointing the command at npx with args -y firebase-tools mcp --dir . (--dir should point at your Firebase project folder).
Restart the MCP client and confirm the Firebase tools appear in its tool list.

.cursor/mcp.json

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

Shown for Cursor. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "firebase": {
      "command": "npx",
      "args": [
        "-y",
        "firebase-tools",
        "mcp",
        "--dir",
        "."
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add firebase -- npx -y firebase-tools mcp --dir .

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

After connecting the firebase MCP server in your client and running firebase login to authenticate, ask "List all of my Firebase projects"; a returned project list confirms the server and authentication are working.

Troubleshooting

Verify Node.js and npm are installed and that `npx firebase-tools` runs successfully on its own
Confirm authentication is complete via `firebase login` or a valid `GOOGLE_APPLICATION_CREDENTIALS` path
Check that the working directory has a valid firebase.json/project alias, or pass `-P <project_id>` explicitly
In CI or headless environments, use a service account or a `firebase login:ci`-generated token instead of interactive login

Things to try

Once connected, you can ask your AI assistant things like:

List all of my Firebase projects
Deploy this project to Firebase Hosting
Show the logs from my deployed Cloud Functions
Export my Firebase Auth users to a data file

Use cases

Letting an AI assistant inspect and deploy Firebase project resources during development

Reading and writing Firestore / Realtime Database data from within a local project directory for debugging

Triggering Hosting or Cloud Functions deployments through an AI assistant

Managing Firebase Auth user import/export or Remote Config templates as part of routine ops work

Supported clients

Cursor

Listed from the project's documentation, not tested by this site.

Overview

The Firebase MCP Server ships inside the firebase/firebase-tools repository and exposes Firebase platform operations over stdio so AI development tools can work with a Firebase project — deploying Hosting sites and Cloud Functions, reading/writing Realtime Database and Firestore data, managing Auth user import/export, and configuring Remote Config or App Distribution. It runs the firebase-tools npm package via npx with the mcp subcommand and relies on whatever Firebase authentication the user has already set up (firebase login, a service account key, or application default credentials).

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 64.1k · Tools 2 Compare with this →

NPM Sentinel MCP Server 79 · B

AI-powered NPM package analysis MCP server

★ 18 · Tools 19 Compare with this →

MartinLoop MCP Server 79 · B

An independent control layer that gives AI coding agents budgets, brakes, and receipts

★ 394 Compare with this →

GitHub MCP Server 76 · B

GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language

★ 34.2k · Tools 15 Compare with this →

Compare with other MCP servers

Source revision 3317c70b216d Data synced 2026-11-04 Read the FMRS scoring method