Amazon Developer

Console

as

Settings
Sign out
Notifications
Alexa
Amazon Appstore
Ring
AWS
Documentation
Support
Contact Us
My Cases

Console

as

Get Started
Design and Develop
Publish
Reference
Support
Skip to main content
AI Assisted Crash Analysis is a feature of the Amazon Devices Builder Tools MCP Server. It lets an AI coding agent analyze and fix your app’s crash bugs. You provide an app crash report (ACR) file and symbol tables (source map files for JavaScript apps and debug libraries for Turbo Modules). The agent then symbolicates the ACR file, detects crashed code, and fixes it if possible.

​
Prerequisites

  • An AI coding IDE with MCP support (Kiro, Cursor, VSCode Copilot, Claude Code, Amazon Q IDE, or Cline)
  • Amazon Devices Builder Tools MCP Server, with its steering doc set up
  • Vega Studio
  • Vega application project

​
Set up workspaces and provide symbol tables

This AI analysis workflow requires symbol tables to symbolicate the ACR file correctly. If you already have them, skip this step. Otherwise, rebuild your app to generate symbol tables, install it, and reproduce the crash to get the ACR file. For example, if your JS app depends on another Turbo Module package, choose Add folder to workspace… in the File menu.

​
Retrieve ACR file

After a crash happens, retrieve the ACR file from the device. Follow the instructions in Retrieve and symbolicate an ACR. If you can reproduce the crash, Vega Studio shows a pop-up when it detects crashes on the device. You can open the ACR file from the pop-up.

​
Start analysis

Open your ACR files in the editor and ask in the AI agent’s chat interface:
If the AI agent can’t find the ACR file, specify its location explicitly. For example: “You can find the ACR file at /path/to/acr-file.acr”. The AI agent automatically finds the necessary symbol tables. If it fails, specify their locations. For example: “You can find source maps in /path/to/source-map.map”. The JS source map is generated at . The Turbo Module’s debug library is generated at . The AI agent invokes the symbolication tool and finds the source of the crash.
  • If the crash occurred in the Vega OS layer, it shows a link to report the issue.
  • If the crash happened in the app layer, the AI agent investigates the reason and proposes a fix. If it doesn’t propose a fix, prompt in the chat interface:
Last modified on May 6, 2026