Skip to content

mcp.md

Latest commit

 

History

History
147 lines (99 loc) · 5.1 KB

mcp.md

File metadata and controls

147 lines (99 loc) · 5.1 KB
title sidebar_position description
Embrace MCP Server
0
Access Embrace mobile observability data through the Model Context Protocol

Embrace MCP Server

The Embrace Model Context Protocol (MCP) server lets you query your mobile app's performance and crash data directly from AI assistants like Claude. You can investigate crashes, monitor app health, and analyze trends without leaving your development workflow.

Embrace's MCP server is available to all Embrace users.

Available tools

Tool Description
list_apps Find and search applications in your Embrace workspace
get_app_details Get health metrics, crash-free rates, and session counts for an app
get_top_versions Identify which app versions are most widely used
list_crashes List top crashes ranked by frequency and user impact
get_crash_details Get detailed information about a specific crash group
get_crash_stack_samples Fetch actual stack traces for crash analysis
get_crash_distribution Break down crash occurrences across a dimension and compare against session baselines

Prerequisites

Before you can use the Embrace MCP server, you need:

  • An active Embrace account with access to at least one application
  • An MCP-compatible AI assistant

Setup examples

Cursor

Follow the MCP Integration instructions in Cursor's official docs.

For example, add to your mcp.json file:

{
  "mcpServers": {
    "embrace-mcp": {
      "url": "https://mcp.embrace.io/mcp"
    }
  }
}

Claude Code

Run:

claude mcp add --transport http embrace https://mcp.embrace.io/mcp

Then start Claude Code and follow the instructions in the /mcp slash command. Reference Anthropic's official docs here for more info. NOTE: Bearer Token authentication is not currently supported.

Opencode

Follow the OAuth MCP integration instructions in Opencode's official docs.

For example, add to your opencode.json file:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "embrace": {
      "type": "remote",
      "url": "https://mcp.embrace.io/mcp"
    }
  }
}

Then authenticate using:

opencode mcp auth embrace

All others

For other AI Assistants supporting MCP, please reference your relevant docs. Official support will come in the near future for these as we continue to improve and add functionality.

The Embrace MCP Server uses a Streamable HTTP transport and is located at https://mcp.embrace.io/mcp.

Common use cases

Daily health check

Check your app's overall health and identify any issues:

  1. Use get_app_details to check application details, session counts, crash-free rates, and more summary statistics
  2. Use get_top_versions to check on app versions ordered by session count
  3. Use list_crashes to identify any new or critical crashes

Example query: "What's the health status of my app today?"

Investigate a critical crash

Deep dive into a specific crash to understand its root cause:

  1. Use list_crashes to find crashes ranked by total occurrences and number of unique users affected
  2. Use get_crash_details to understand the crash's scope and affected versions
  3. Use get_crash_stack_samples to analyze actual stack traces by crash group ID and identify the bug

Example query: "What's causing the top crash in my app and how do I fix it?"

Investigate crash distribution

Use get_crash_distribution to see where a crash is concentrated by comparing its occurrence across a dimension against the session baseline.

  1. Use list_crashes to find a crash group ID
  2. Use get_crash_distribution with group_by: "os_version" to see which OS versions are over-represented relative to their session share
  3. Use get_crash_distribution with group_by: "model" or group_by: "country" to further narrow down affected segments

Example queries:

  • "Which device models are most affected by this crash?"
  • "Is this crash happening more often on specific OS versions?"
  • "Is this crash concentrated in a particular country?"

Version-specific analysis

Focus on a particular app version:

  1. Use list_crashes with the app_versions parameter to filter results
  2. Use get_crash_details for the top crashes in that version
  3. Use get_crash_stack_samples to see if patterns emerge

Example query: "Show me all crashes in version 2.1.0"

Track trends over time

Monitor how your app's health changes:

  1. Use get_app_details with different time_window values to compare periods
  2. Use list_crashes to see which crashes are new or resolved
  3. Use get_top_versions to track version adoption velocity

Example query: "How has my crash-free rate changed over the last 7 days?"