AI assistance for Crashlytics via MCP

With Firebase Crashlytics MCP tools and prompts, you can interact with your Crashlytics data using your MCP-capable AI-powered development tools, like Gemini CLI, Claude Code, or Cursor. These Crashlytics MCP tools and prompts provide critical context to your AI tool to help you manage, prioritize, debug, and fix issues.

After you set up the Firebase MCP server, you can use the MCP features for Crashlytics:

Guided workflow to prioritize and fix issues

Free-form conversational debugging

info What is MCP? Model Context Protocol (MCP) is a standardized way for AI tools to access external tools and data sources.

Before you begin: Set up the Firebase MCP server

This section describes the basic setup of the Firebase MCP server so that you can use Crashlytics-specific MCP commands, prompts, and tools (which are all described later on this page).

Prerequisites

Make sure your environment meets these requirements:

• A working installation of Node.js and npm. Installing Node.js automatically installs the npm command tools.

• Your AI-powered development tool supports MCP integrations.

If you're working with a Unity project, review the FAQ about loading Crashlytics MCP tools and prompts.

Configure your AI tool to use the Firebase MCP server

gemini extensions install https://github.com/gemini-cli-extensions/firebase/

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

gemini extensions install https://github.com/gemini-cli-extensions/firebase/

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

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

claude mcp add firebase npx -- -y firebase-tools@latest mcp

claude mcp list

firebase: npx -y firebase-tools@latest mcp - ✓ Connected

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

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

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

"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"]
  }
}

In addition to the basic configuration described above for each AI tool, you can specify optional parameters.

(Recommended) Guided workflow to prioritize and fix issues with crashlytics:connect

Crashlytics provides a guided workflow that's conversational and flexible to help you prioritize and fix Crashlytics issues in your app. For example, your AI tool can fetch issues, explain issues, identify potential fixes, and even make the code changes for you.

This guided workflow is available via the crashlytics:connect MCP command.

Access and use the command

1. If you haven't already, set up the Firebase MCP server, and then start your AI tool.

2. Run the crashlytics:connect MCP command.

   Most AI tools provide a way to conveniently access this workflow. For example, if you're using the Gemini CLI, run the slash command /crashlytics:connect

3. Use your AI tool to help prioritize and fix Crashlytics issues, for example:

   • View a list of prioritized issues.
   • Debug a specific issue by providing its ID or URL.
   • Request more information about a crash.
   • Ask the agent its reasoning for a suggested root cause.

Free-form conversational debugging

While we recommend using crashlytics:connect for the best debugging experience, you can also debug issues using a free-form conversation with an AI tool that has access to the Crashlytics MCP tools. This is especially important for AI tools that don't yet support MCP prompts (often referred to as slash commands or custom commands).

After you've set up the Firebase MCP server, try some of the following examples.

Fetch an issue and crash context

When your AI tool has access to Crashlytics MCP tools, it can fetch critical Crashlytics issue data like user and event counts, stacktraces, metadata, and app version information.

Here are some example prompts:

• A customer reported an issue during login when using our latest release. What Crashlytics issues do I have that could be related to this login trouble?

  • To answer this question, your AI tool will likely read your code to understand where login happens and use various Crashlytics MCP tools to retrieve issue data. Your AI tool will then try to determine whether an issue exists in the latest version that relates to the login flow.

• The previous on-call engineer was investigating issue abc123 but wasn't able to resolve it. She said she left some notes -- let's pick up where she left off.

  • To answer this question, your AI tool will use various Crashlytics MCP tools to retrieve issue context and any notes posted to the issue. It might also fetch example crashes to resume the investigation into the issue.

Document a debugging investigation

When debugging an issue, it's often helpful to maintain records for yourself or your team. Crashlytics offers this capability in the Firebase console, and your AI tool equipped with Crashlytics MCP tools can also help — for example: summarize an investigation, add a note with helpful metadata (like a link to a Jira or GitHub issue), or close an issue after it's fixed.

Here are some example prompts:

• Add a note to issue abc123 summarizing this investigation and the proposed fix.
• We weren't able to get to the bottom of this issue today, summarize what we learned and attach it to issue abc123 to pick back up later.
• Close issue abc123 and leave a note including the link to the PR that fixed the issue.

Crashlytics MCP tool reference

The following tables list the Crashlytics MCP tools that are available via the Firebase MCP server.

After you've set up the Firebase MCP server, your AI tool can use these MCP tools to help you understand, debug, and manage issues. These MCP tools are used in both the crashlytics:connect guided workflow and in free-form conversations with your AI tool.

For the majority of use cases, these MCP tools are for LLM-use only and not for direct use by a human developer. The LLM decides when to use these MCP tools based on your interaction with your AI tool.

Manage Crashlytics issues

The following table describes the tools available to manage your Crashlytics issues.

Fetch Crashlytics data

The following table describes the tools available to get Crashlytics-related information about your apps.

Additional information

How your data is used

Data governance is determined by the AI-powered development tool that you use, and is subject to the terms defined by that AI tool.

Pricing

Firebase doesn't charge you to use Crashlytics MCP tools and prompts or to fetch Crashlytics data from our public API.

Any cost is determined by the AI-powered development tool that you use, and could be determined by the volume of Crashlytics data that's used by the AI tool. Note that Firebase doesn't offer an explicit way to control how much data is loaded into context, but we do include sensible default guidance for the model.

Troubleshooting and FAQ