Microsoft Amplifier

"I have more ideas than time to try them out" — The problem we're solving

Amplifier is a complete development environment that takes AI coding assistants and supercharges them with discovered patterns, specialized expertise, and powerful automation — turning a helpful assistant into a force multiplier that can deliver complex solutions with minimal hand-holding.

We've taken our learnings about what works in AI-assisted development and packaged them into a ready-to-use environment. Instead of starting from scratch every session, you get immediate access to proven patterns, specialized agents for different tasks, and workflows that actually work.

Amplifier provides powerful tools and systems:

• 20+ Specialized Agents: Each expert in specific tasks (architecture, debugging, security, etc.)
• Pre-loaded Context: Proven patterns and philosophies built into the environment
• Parallel Worktree System: Build and test multiple solutions simultaneously
• Knowledge Extraction System: Transform your documentation into queryable, connected knowledge
• Conversation Transcripts: Never lose context - automatic export before compaction, instant restoration
• Automation Tools: Quality checks and patterns enforced automatically

Before starting, you'll need:

• Python 3.11+ - Download Python
• UV - Install UV
• Node.js - Download Node.js
• VS Code (recommended) - Download VS Code
• Git - Download Git

Platform Note: Development and testing has primarily been done in Windows WSL2. macOS and Linux should work but have received less testing. Your mileage may vary.

1. Clone the repository:

   git clone https://github.com/microsoft/amplifier.git cd amplifier

2. Run the installer:

   This installs Python dependencies, the Claude CLI, and sets up your environment.

3. Configure your data directories (Recommended but optional):

   Why configure this? By default, Amplifier stores data in .data/ (git-ignored). But centralizing your data externally gives you:

   • Shared knowledge across all worktrees - Every parallel experiment accesses the same knowledge base
   • Cross-device synchronization - Work from any machine with the same accumulated knowledge
   • Automatic cloud backup - Never lose your extracted insights
   • Reusable across projects - Apply learned patterns to new codebases

   Set up external directories:

   cp .env.example .env # Edit .env to point to your preferred locations

   Example configuration using cloud storage:

   # Centralized knowledge base - shared across all worktrees and devices # Using OneDrive/Dropbox/iCloud enables automatic backup! AMPLIFIER_DATA_DIR=~/OneDrive/amplifier/data # Your source materials (documentation, specs, design docs, notes) # Can point to multiple folders where you keep content AMPLIFIER_CONTENT_DIRS=.data/content,~/OneDrive/amplifier/content,~/Documents/notes

4. Activate the environment (if not already active):

   source .venv/bin/activate # Linux/Mac/WSL .venv\Scripts\activate # Windows

Start Claude in the Amplifier directory to get all enhancements automatically:

Want Amplifier's power on your own code? Easy:

1. Start Claude with both directories:

   claude --add-dir /path/to/your/project

2. Tell Claude where to work (paste as first message):

   I'm working in /path/to/your/project which doesn't have Amplifier files. Please cd to that directory and work there. Do NOT update any issues or PRs in the Amplifier repo.

3. Use Amplifier's agents on your code:

   • "Use the zen-architect agent to design my application's caching layer"
   • "Deploy bug-hunter to find why my login system is failing"
   • "Have security-guardian review my API implementation for vulnerabilities"

Why use this? Stop wondering "what if" — build multiple solutions simultaneously and pick the winner.

Each worktree is completely isolated with its own branch, environment, and context.

See the Worktree Guide for advanced features, such as hiding worktrees from VSCode when not in use, adopting branches from other machines, and more.

See costs, model, and session info at a glance:

Example: ~/repos/amplifier (main → origin) Opus 4.1 💰$4.67 ⏱18m

Shows:

• Current directory and git branch/status
• Model name with cost-tier coloring (red=high, yellow=medium, blue=low)
• Running session cost and duration

Enable with:

Want to get the most out of Amplifier? Check out The Amplifier Way for battle-tested strategies including:

• Understanding capability vs. context
• Decomposition strategies for complex tasks
• Using transcript tools to capture and improve workflows
• Demo-driven development patterns
• Practical tips for effective AI-assisted development

Instead of one generalist AI, you get 20+ specialists:

Core Development:

• zen-architect - Designs with ruthless simplicity
• modular-builder - Builds following modular principles
• bug-hunter - Systematic debugging
• test-coverage - Comprehensive testing
• api-contract-designer - Clean API design

Analysis & Optimization:

• security-guardian - Security analysis
• performance-optimizer - Performance profiling
• database-architect - Database design and optimization
• integration-specialist - External service integration

Knowledge & Insights:

• insight-synthesizer - Finds hidden connections
• knowledge-archaeologist - Traces idea evolution
• concept-extractor - Extracts knowledge from documents
• ambiguity-guardian - Preserves productive contradictions

Meta & Support:

• subagent-architect - Creates new specialized agents
• post-task-cleanup - Maintains codebase hygiene
• content-researcher - Researches from content collection

[See .claude/AGENTS_CATALOG.md for the complete list]

Why use this? Stop losing insights. Every document, specification, design decision, and lesson learned becomes part of your permanent knowledge that Claude can instantly access.

1. Add your content (any text-based files: documentation, specs, notes, decisions, etc.)

2. Build your knowledge base:

   make knowledge-update # Extracts concepts, relationships, patterns

3. Query your accumulated wisdom:

   make knowledge-query Q="authentication patterns" make knowledge-graph-viz # See how ideas connect

Never lose context again. Amplifier automatically exports your entire conversation before compaction, preserving all the details that would otherwise be lost. When Claude Code compacts your conversation to stay within token limits, you can instantly restore the full history.

Automatic Export: A PreCompact hook captures your conversation before any compaction event:

• Saves complete transcript with all content types (messages, tool usage, thinking blocks)
• Timestamps and organizes transcripts in .data/transcripts/
• Works for both manual (/compact) and auto-compact events

Easy Restoration: Use the /transcripts command in Claude Code to restore your full conversation:

The transcript system helps you:

• Continue complex work after compaction without losing details
• Review past decisions with full context
• Search through conversations to find specific discussions
• Export conversations for sharing or documentation

Transcript Commands (via Makefile):

A one-command workflow to go from an idea to a module (Contract & Spec → Plan → Generate → Review) inside the Amplifier Claude Code environment.

• Run inside a Claude Code session:
  /modular-build Build a module that reads markdown summaries, synthesizes net-new ideas with provenance, and expands them into plans. mode: auto level: moderate
• Docs: see docs/MODULAR_BUILDER_LITE.md for the detailed flow and guardrails.
• Artifacts: planning goes to ai_working/<module>/… (contract/spec/plan/review); code & tests to amplifier/<module>/….
• Isolation & discipline: workers read only this module’s contract/spec plus dependency contracts. The spec’s Output Files are the single source of truth for what gets written. Every contract Conformance Criterion maps to tests. 〔Authoring Guide〕

• auto (default): runs autonomously if confidence ≥ 0.75; otherwise falls back to assist.
• assist: asks ≤ 5 crisp questions to resolve ambiguity, then proceeds.
• dry-run: plan/validate only (no code writes).

Re‑run /modular-build with a follow‑up ask; it resumes from ai_working/<module>/session.json.

1. Design: "Use zen-architect to design my notification system"
2. Build: "Have modular-builder implement the notification module"
3. Test: "Deploy test-coverage to add tests for the new notification feature"

1. Investigate: "Use bug-hunter to find why my application's API calls are failing"
2. Verify: "Have security-guardian review my authentication implementation"

1. Extract: make knowledge-update (processes your documentation)
2. Query: make knowledge-query Q="error handling patterns"
3. Apply: "Implement error handling using patterns from our knowledge base"

Want to create tools like the ones in the scenarios/ directory? You don't need to be a programmer.

Not sure what to build? Ask Amplifier to brainstorm with you:

This brainstorming session will give you ideas like:

• Documentation Quality Amplifier - Improves docs by simulating confused readers
• Research Synthesis Quality Escalator - Extracts and refines knowledge from documents
• Code Quality Evolution Engine - Writes code, tests it, learns from failures
• Multi-Perspective Consensus Builder - Simulates different viewpoints to find optimal solutions
• Self-Debugging Error Recovery - Learns to fix errors autonomously

The magic happens when you combine:

1. Amplifier's brainstorming - Generates diverse possibilities
2. Your domain knowledge - You know your needs and opportunities
3. Your creativity - Sparks recognition of what would be useful

Once you have an idea:

1. Describe your goal - What problem are you solving?
2. Describe the thinking process - How should the tool approach it?
3. Let Amplifier build it - Use /ultrathink-task to create the tool
4. Iterate to refine - Provide feedback as you use it
5. Share it back - Help others by contributing to scenarios/

Example: The blog writer tool was created with one conversation where the user described:

• The goal (write blog posts in my style)
• The thinking process (extract style → draft → review sources → review style → get feedback → refine)

No code was written by the user. Just description → Amplifier builds → feedback → refinement.

For detailed guidance, see scenarios/blog_writer/HOW_TO_CREATE_YOUR_OWN.md.

[!IMPORTANT] > This is an experimental system. We break things frequently.

• Not accepting contributions yet (but we plan to!)
• No stability guarantees
• Pin commits if you need consistency
• This is a learning resource, not production software
• No support provided - See SUPPORT.md

We're building toward a future where:

1. You describe, AI builds - Natural language to working systems
2. Parallel exploration - Test 10 approaches simultaneously
3. Knowledge compounds - Every project makes you more effective
4. AI handles the tedious - You focus on creative decisions

The patterns, knowledge base, and workflows in Amplifier are designed to be portable and tool-agnostic, ready to evolve with the best available AI technologies.

See AMPLIFIER_VISION.md for details.

• Knowledge extraction works best in Claude environment
• Processing time: ~10-30 seconds per document
• Memory system still in development

"The best AI system isn't the smartest - it's the one that makes YOU most effective."

Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit Contributor License Agreements.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.