Build AI Your Way
With Amplifier
Pick your model. Add your tools. Define your workflow. Every piece works independently, swaps instantly, and composes into something powerful.
uv tool install git+https://github.com/microsoft/amplifierAmplifier is open by design.
Most AI tools hide how they work. Amplifier shows you everything. Inspect any component, swap any piece, own the whole stack.
Use any model
The model is a component, not a cage. Switch providers without rewriting anything. Your tools and workflows just work.
See everything
Watch it think. Every decision logged, every tool call visible. When something breaks, you know exactly where and why.
Own your setup
Your agent is a file you can read, version, and share. Not a subscription. Not a black box. Yours.
Built from composable pieces.
Providers connect to any LLM. Tools give agents capabilities. Workflows orchestrate complex tasks. Each piece works independently, but they compose into something powerful. That's why it's a platform, not just a wrapper around someone else's API.
Providers
Connect to any LLM. Anthropic, OpenAI, Google, Azure, AWS, Ollama, or bring your own.
Tools
Give your agent capabilities. File access, web search, code execution, or build your own.
Agents
Specialized personas for specific tasks. Security reviewers, writers, architects, analysts.
Recipes
Multi-step workflows with checkpoints. Chain agents together, add approval gates, resume anytime.
See how it works.
Run one command and watch the entire flow. Every tool call is logged. Every decision is explained. You're not looking at a black box that gives you answers - you're watching a transparent system that shows you how it thinks.
Capture your workflow as code.
A bundle is your entire setup in plain text. The providers you trust, the tools you need, the behaviors you've refined. As you work, you adapt it. As you share it, others adapt theirs. The more you use it, the more it becomes yours.
---
bundle:
name: security-reviewer
version: 1.0.0
providers:
- anthropic # Or openai, google, ollama...
tools:
- filesystem # Read code
- search # Find patterns
- bash # Run security tools
behaviors:
- secure-by-default # Security-first mindset
---
# Security Reviewer
You review code for vulnerabilities.
## Focus on:
- SQL injection, XSS, auth bypasses
- Secrets in code
- Missing input validation42 lines. That's the entire agent.
Debug in minutes
Every tool call logged. When it breaks, trace the exact line that failed.
Swap anything
Change models, tools, providers. Your setup evolves as fast as AI does.
Share like code
Git push. PR review. Hand it to a teammate. No screenshots, no wikis.
Your data stays yours
Run locally. Inspect every call. Add approval gates. Nothing leaves without your say.
Built by different minds.
Ready for yours.
Every bundle carries someone's approach to solving problems. Install theirs. Add your perspective. When you share back, you're teaching others a new way to think. The more perspectives, the smarter the community.
Recipes
Multi-step AI agent orchestration with checkpoints and approval gates
Design Intelligence
Specialized agents for design systems, components, and UI work
Issues
Issue management and triage workflows
Transcribe
Transform YouTube videos and audio files into searchable transcripts with AI-powered insights
Blog Creator
AI-powered blog creation with style-aware generation and rich markdown editor
Voice Assistant
Desktop voice assistant with native speech-to-speech via OpenAI Realtime API
Tool Generator
AI-powered tool generator for creating custom Amplifier tools
Playground
Interactive environment for building, configuring, and testing Amplifier AI agent sessions
Session Analyzer
Analyze Amplifier session logs and generate usage metrics and insights
Benchmarks
Benchmarking and evaluation suite for Amplifier performance
Lakehouse
Amplifier on top of your data - daemon and webapp for data integration
AWS Bedrock
AWS Bedrock integration with cross-region inference support for Claude models
Recipes
Multi-step AI agent orchestration with checkpoints and approval gates
Design Intelligence
Specialized agents for design systems, components, and UI work
Issues
Issue management and triage workflows
Transcribe
Transform YouTube videos and audio files into searchable transcripts with AI-powered insights
Blog Creator
AI-powered blog creation with style-aware generation and rich markdown editor
Voice Assistant
Desktop voice assistant with native speech-to-speech via OpenAI Realtime API
Tool Generator
AI-powered tool generator for creating custom Amplifier tools
Playground
Interactive environment for building, configuring, and testing Amplifier AI agent sessions
Session Analyzer
Analyze Amplifier session logs and generate usage metrics and insights
Benchmarks
Benchmarking and evaluation suite for Amplifier performance
Lakehouse
Amplifier on top of your data - daemon and webapp for data integration
AWS Bedrock
AWS Bedrock integration with cross-region inference support for Claude models
Recipes
Multi-step AI agent orchestration with checkpoints and approval gates
Design Intelligence
Specialized agents for design systems, components, and UI work
Issues
Issue management and triage workflows
Transcribe
Transform YouTube videos and audio files into searchable transcripts with AI-powered insights
Blog Creator
AI-powered blog creation with style-aware generation and rich markdown editor
Voice Assistant
Desktop voice assistant with native speech-to-speech via OpenAI Realtime API
Tool Generator
AI-powered tool generator for creating custom Amplifier tools
Playground
Interactive environment for building, configuring, and testing Amplifier AI agent sessions
Session Analyzer
Analyze Amplifier session logs and generate usage metrics and insights
Benchmarks
Benchmarking and evaluation suite for Amplifier performance
Lakehouse
Amplifier on top of your data - daemon and webapp for data integration
AWS Bedrock
AWS Bedrock integration with cross-region inference support for Claude models
OpenAI Realtime
OpenAI Realtime API for native speech-to-speech with ultra-low latency
Memory
Persistent memory system with automatic capture and progressive disclosure
Expert Cookbook
Achieve state-of-the-art results with best practices and patterns
Image Generation
Multi-provider AI image generation with DALL-E, Imagen, and GPT-Image-1
YouTube Download
Download audio and video from YouTube with metadata extraction
Whisper
Speech-to-text transcription using OpenAI Whisper API
Style Extraction
Extract and apply writing styles from text samples
Markdown Utils
Markdown parsing, injection, and metadata extraction utilities
RLM
Process 10M+ token contexts via Recursive Language Model and sandboxed REPL
Tool Memory
Persistent memory tool for storing and retrieving facts across sessions
Event Broadcast
Transport-agnostic event broadcasting for streaming UI applications
Explanatory Mode
Inject explanatory output style with educational Insight blocks
OpenAI Realtime
OpenAI Realtime API for native speech-to-speech with ultra-low latency
Memory
Persistent memory system with automatic capture and progressive disclosure
Expert Cookbook
Achieve state-of-the-art results with best practices and patterns
Image Generation
Multi-provider AI image generation with DALL-E, Imagen, and GPT-Image-1
YouTube Download
Download audio and video from YouTube with metadata extraction
Whisper
Speech-to-text transcription using OpenAI Whisper API
Style Extraction
Extract and apply writing styles from text samples
Markdown Utils
Markdown parsing, injection, and metadata extraction utilities
RLM
Process 10M+ token contexts via Recursive Language Model and sandboxed REPL
Tool Memory
Persistent memory tool for storing and retrieving facts across sessions
Event Broadcast
Transport-agnostic event broadcasting for streaming UI applications
Explanatory Mode
Inject explanatory output style with educational Insight blocks
OpenAI Realtime
OpenAI Realtime API for native speech-to-speech with ultra-low latency
Memory
Persistent memory system with automatic capture and progressive disclosure
Expert Cookbook
Achieve state-of-the-art results with best practices and patterns
Image Generation
Multi-provider AI image generation with DALL-E, Imagen, and GPT-Image-1
YouTube Download
Download audio and video from YouTube with metadata extraction
Whisper
Speech-to-text transcription using OpenAI Whisper API
Style Extraction
Extract and apply writing styles from text samples
Markdown Utils
Markdown parsing, injection, and metadata extraction utilities
RLM
Process 10M+ token contexts via Recursive Language Model and sandboxed REPL
Tool Memory
Persistent memory tool for storing and retrieving facts across sessions
Event Broadcast
Transport-agnostic event broadcasting for streaming UI applications
Explanatory Mode
Inject explanatory output style with educational Insight blocks
Ready to build?
uv tool install git+https://github.com/microsoft/amplifierQuestions? Join the discussion





