Set Up Amazon Devices Builder Tools for AI-Powered Development
Amazon Devices Builder Tools for AI-Powered Development includes a Model Context Protocol (MCP) Server and Agent Skills. These components provide Amazon Devices-specific metadata, tools, and workflows to your AI coding agent, so your agent follows recommended patterns and best practices for project setup, feature integrations, performance analysis, crash debugging, and testing.
Amazon Devices Builder Tools provides support for Vega OS, which powers Fire TV. For developers also building on Fire OS, it supports In-App Purchase (IAP) integration and app migration workflows.
Amazon Devices Builder Tools includes:
- Builder Tools MCP Server — A standalone npm package that provides Amazon Devices metadata, tools, and capabilities to compatible AI coding agents.
- Builder Tools Agent Skills — Lightweight, modular, markdown-based instruction sets that guide your agent through Fire TV development workflows.
- Steering documents — Context files that ground your AI assistant in Amazon Devices best practices.
Amazon Devices Builder Tools doesn't collect, transmit, or store any of your confidential information. For details, see Privacy FAQ.
Amazon Devices Builder Tools is a standalone npm package and isn't bundled with the Vega SDK. Install it separately using the steps on this page. For the full list of features, supported agents, manual configuration options, and tool reference, see the Amazon Devices Builder Tools ReadMe.
@amazon-devices/amazon-devices-buildertools-mcp) was previously called Vega DevTools (@amazon-devices/vega-devtools-mcp). If you have the previous version installed, delete @amazon-devices/vega-devtools-mcp from the config file and add @amazon-devices/amazon-devices-buildertools-mcp to receive the latest updates and features.Prerequisites
- Node.js 18 or later
- Vega SDK v0.22 or later (required only for Vega OS development workflows)
Get started
Run the following set-up command in your terminal (Bash or ZSH shell) to configure Amazon Devices Builder Tools for your AI agent and add Amazon Devices-specific context to your project:
npx -y @amazon-devices/amazon-devices-buildertools-mcp@latest init-context
This command installs and configures MCP, Steering, and Skills for your AI agent. It prompts you to select your agent, choose an installation directory, and handle existing context files. On subsequent runs of init-context, the command checks for updates and downloads new capabilities automatically.
When the command completes successfully, you see output similar to:
✅ MCP configuration updated
✅ Context document installed
✅ Setup complete for Kiro
Interactive mode (default)
When running in interactive mode, you're guided through:
- Select your AI agent (for example, Cursor, Kiro, or Other/Custom).
- Enter
y/nto automatically update your agent's MCP settings file. - Choose the default context document installation path or enter a custom path.
- Choose action:
merge(add new content to existing file),update(overwrite existing file), orsave(write to a new file).
Non-interactive mode
Use command-line options to skip prompts and automate the setup process:
npx -y @amazon-devices/amazon-devices-buildertools-mcp@latest init-context --agent <agent-name> --force
Available options:
| Option | Description |
|---|---|
-a, --agent <type> |
Target AI agent (cursor, cline, kiro, claude-code-cli, claude-code-desktop, github-copilot, amazon-q, other) |
-p, --context-document-path |
Path to save context document |
-d, --skip-context-document |
Skip context document installation (MCP config only) |
-m, --skip-mcp-config |
Skip MCP configuration (context document only) |
-f, --force |
Skip all confirmation prompts and force overwrite |
-h, --help |
Show help for init-context command |
Check agent setup status
After installing the MCP server and context documents, verify your setup using the check-status command:
npx -y @amazon-devices/amazon-devices-buildertools-mcp@latest check-status
To check a specific agent:
npx -y @amazon-devices/amazon-devices-buildertools-mcp@latest check-status --agent <agent-name>
This command checks:
- Context document installation status and version
- MCP configuration status
- Detection of legacy configurations
Example output:
Agent │ Context Document │ MCP Configuration
───────────────────────┼──────────────────────────┼──────────────────────────
Kiro │ ✅ v5.1 │ ✅ Configured
📊 Summary:
Total agents checked: 1
✅ Fully configured: 1
Verify setup
After installation, first run check-status to confirm your configuration is correct (see Check agent setup status). Then confirm your agent can access Amazon Devices Builder Tools. In your AI agent's chat, enter:
List the tools provided by Amazon Devices Builder Tools and Agent Skills
Your agent should respond with a list of available tools including analyze_perfetto_traces, get_app_hot_functions, list_documents, read_asset, read_document, search_documentation, and symbolicate_acr.
Then try an Amazon Devices-specific prompt:
Help me setup Vega SDK
If your agent responds with Amazon Devices-specific setup steps (rather than generic React Native instructions), the configuration is working correctly. If you encounter issues, see Troubleshoot Amazon Devices Builder Tools for AI-Powered Development.
Once setup is verified, try prompts like "Build and run my app on Virtual Device" or "Help me debug TTFF in my app" to use the full range of Amazon Devices Builder Tools capabilities. See Supported workflows for more examples.
Builder Tools Agent Skills
Builder Tools Agent Skills are lightweight, modular, markdown-based (SKILL.md) instruction sets that guide your agent through verified steps for common Fire TV development and debugging workflows. Agent Skills trigger automatically when your prompt matches the skill's metadata.
The following skills are available:
| Skill | Purpose |
|---|---|
| amazon-devices-vega-setup-sdk | To install SDK or CLI commands |
| amazon-devices-vega-build-and-run | To build, install, run, or deploy app |
| amazon-devices-vega-navigation | To implement screen navigation |
| amazon-devices-vega-focus-management | To handle D-Pad/remote focus |
| amazon-devices-vega-media-player | To implement video/audio playback |
| amazon-devices-vega-ui-components | For UI components (buttons, lists, cards) |
| amazon-devices-vega-app-manifest | To configure manifest.toml |
| amazon-devices-vega-app-performance | To debug or optimize app performance |
| amazon-devices-vega-best-practices | For development best practices |
| vega-multi-tv-migration | To migrate Vega apps to a cross-device React Native monorepo including Fire OS |
These skills work alongside any other skills you use or create. New skills are added based on developer feedback.
MCP tools
The Builder Tools MCP server provides the following tools to assist with Vega app development:
| Tool | Description |
|---|---|
| analyze_perfetto_traces | Analyze Vega platform traces using Perfetto trace processor to extract KPI metrics and performance data. Helps diagnose performance issues and analyze app launch times (TTFF, TTFD). |
| get_app_hot_functions | Read and analyze CPU trace files to identify hot functions (CPU-intensive operations) contributing to performance bottlenecks. Helps pinpoint which functions consume the most CPU time. |
| list_documents | List all available Vega documents related to app development for Amazon Vega OS. Returns name and description of available documents. Supports filtering by type (KB, PROMPT, STEERING, WORKFLOW, SKILL). |
| read_asset | Retrieve assets referenced in agentic workflows. Assets are saved to a temporary location. Supports images and scripts. |
| read_document | Read documents related to app development for Amazon Vega OS. Provides access to comprehensive documentation about Vega app development and debugging topics. |
| search_documentation | Search Vega development documentation and agentic workflows using full-text search. Returns document names and relevant snippets matching the query. |
| symbolicate_acr | Symbolicate ACR (Amazon Crash Report) files and create a report. Supports both GDB-based native crash symbolication and JavaScript stack trace symbolication. |
MCP prompts
Builder Tools MCP provides the following pre-defined prompt templates for common workflows. Execute them using /prompts if your AI agent supports MCP Prompts:
| Prompt | Description |
|---|---|
| apply_performance_best_practices | Diagnose and optimize React Native application performance issues including component rendering, memory management, navigation, network optimization, and state management. |
| detect_component_re-renders | Diagnose and optimize Vega application UI fluidity performance issues caused by component re-rendering using React Native tools. |
| diagnose-slowAppLaunch | Diagnose Vega application's slow app launch KPIs (TTFF and/or TTFD) using a KPI report file. |
| diagnose_crash | Diagnose crashes in Vega applications including JavaScript, Native, LMK, and ANR crashes. Automatically discovers ACR files, symbolicates stack traces, and routes to the appropriate analysis workflow. |
| diagnose_key_input_latency | Diagnose and resolve key input latency issues in React Native Vega apps through systematic analysis of JS Thread performance, CPU bottlenecks, and input event handling optimization. |
| diagnose_ui_fluidity | Diagnose UI fluidity KPI failures in React Native Vega apps through systematic analysis of component re-rendering and CPU performance bottlenecks. |
| fix_hot_functions | Diagnose and optimize Vega application UI fluidity issues by analyzing and fixing CPU-intensive hot functions. |
| react_native_for_vega_add_simple_media_playback_implementation | Add a simple media playback implementation to an existing React Native for Vega application. |
| upgrade_carousel_component | Assist in migrating to newer versions of the Carousel component in the Vega SDK. |
Supported workflows
In addition to Agent Skills and MCP Prompts, Amazon Devices Builder Tools support the following workflows. For the full list of supported workflows, see the Amazon Devices Builder Tools ReadMe.
| Development area | Description | Sample prompts |
|---|---|---|
| Onboarding | Set up your environment for Vega app development | "Help me setup Vega SDK", "Help me set up Vega app project", "Tell me about Vega app development" |
| App Porting (Beta) | Port Fire OS apps to Vega or wrap a web URL as a Vega app, with auto-detection of app type | "Port my TV app to Vega", "Convert my Fire OS WebView app to a Vega Web App", "Port my Fire OS React Native app to VegaScript" |
| Feature integrations | Integrate and test Amazon Appstore In-App Purchasing (IAP) SDK in Vega and Fire OS apps | "Help me integrate IAP in my Vega app", "How do I set up In-App Purchasing for my Vega app?", "Help me test IAP in my Vega app" |
| UI Development | Create Vega app UI from screenshots and design mockups | "I have an image at {image_path}. Create a Vega app to match the image." |
| Focus Management | Implement D-Pad navigation and focus management for TV interfaces | "Can you create an app with 2 buttons to show case focus management?" |
| Give Feedback | Provide feedback to Amazon about your experience with the MCP | "I want to provide feedback", "How can I provide vega mcp feedback?" |
| Performance - responsiveness | Debug and fix Time to First Frame (TTFF), Time to Fully Drawn (TTFD), UI re-rendering, and Input Latency | "Can you help me debug and fix the TTFF issue in my app?", "Can you help me minimize unnecessary re-renders in my app?", "Why is my app responding slowly to key presses?" |
| Performance - fluidity | Diagnose and fix fluidity failures through component re-rendering and CPU performance analysis | "Help me fix frame drops and jank in my Vega app", "Why is scrolling not smooth in my app?" |
| Performance best practices | Analyze components and code to ensure they follow React Native Vega performance best practices | "Help me analyze if this component is following React Native Vega performance best practices" |
| Crash Analysis | Diagnose JavaScript, Native, LMK, and ANR crashes with automated ACR analysis | "Why did my app crash?", "Help me analyze this ACR file /path/to/crash.acr" |
| Key Input Latency | Diagnose and resolve key input latency issues through JS Thread performance and input event analysis | "Why is my app responding slowly to key presses?", "Help me diagnose key input latency in my Vega app" |
| UI Fluidity | Diagnose UI fluidity KPI failures through component re-rendering and CPU performance analysis | "My app's UI fluidity score is low, help me diagnose it", "Why is scrolling not smooth in my app?" |
| SDK & CLI | Install/update SDK and run CLI tools via spec-driven CLI | "Install/update SDK from AI agents", "Build my app", "Run my app on virtual device" |
| Remote Knowledge Search | Search and retrieve documentation from developer.amazon.com and community.amazondeveloper.com to get AI assistance on any documented app development topic | "How can I do a beta test of my Vega app?", "What are the accessibility guidelines for Vega apps?", "How do I submit my app to the Amazon Appstore?" |
Privacy FAQ
Q1: Does Amazon Devices Builder Tools collect or send my code, queries, or other confidential data to Amazon?
A: No. Amazon Devices Builder Tools doesn't collect, transmit, or store any of your confidential information, including:
- Your code samples or app source code.
- Your queries or prompts to the AI assistant.
- Your project files or configurations.
- Any other developer data or intellectual property.
However, documentation search queries issued by your AI agent are sent to Amazon Services to find relevant documentation for the task at hand.
All interactions between Amazon Devices Builder Tools and your AI assistant happen locally on your machine. Your code and queries remain private and are processed only by your chosen AI assistant, according to that assistant's own privacy policies. The only exception is documentation search queries.
Q2: What data does the AI assistant have access to?
A: Your AI assistant processes your queries and code according to its own privacy policy and terms of service. Amazon Devices Builder Tools provides tools and Amazon Devices-specific context to your assistant. The assistant's data handling practices are governed by its provider. Review your AI assistant's privacy policy to understand how it handles your data.
Q3: Will Amazon collect any data about my use of Amazon Devices Builder Tools?
A: Amazon collects anonymous telemetry data to improve Amazon Devices Builder Tools. This telemetry includes feature usage patterns, error rates, and performance metrics. You can disable telemetry collection in the settings.
Important privacy protections:
- No confidential data is collected: Your queries, code samples, project files, or any other developer-created content are never sent to Amazon.
- Local processing: The MCP runs entirely on your local machine. The only exception is documentation search queries (see Q1).
- Anonymous data only: Telemetry data doesn't include personally identifiable information.
Amazon also offers an optional developer survey to understand the usefulness of Amazon Devices Builder Tools. Participation in this survey is voluntary.
Troubleshooting
If you encounter any issues, see Troubleshoot Amazon Devices Builder Tools for AI-Powered Development.
Related topics
- Amazon Devices Builder Tools ReadMe
- Build your app
- Run your app on Vega Virtual Device or Fire TV Stick
- App Performance Best Practices
- Debug Crashes
Last updated: May 20, 2026
