Amplifier

Amplifier

30+ specialized AI agents and a knowledge management system for your codebase.

Overview

Amplifier extends the IDE with specialized AI agents, each trained for specific tasks. Instead of one general-purpose AI, you get experts in architecture, security, testing, and more. Amplifier also maintains a knowledge graph of your codebase for context-aware assistance.

AI Agents

Built-in Agents (30+)

Agents are organized by category:

Analysis & Debugging (4 agents)

AgentDescription
bug-hunterFinds and diagnoses bugs
code-analyzerStatic analysis and patterns
performance-profilerPerformance bottleneck detection
error-tracerStack trace analysis

Architecture & Design (6 agents)

AgentDescription
zen-architectHolistic system design
api-designerREST/GraphQL API design
database-architectSchema and query optimization
security-architectSecurity-first design
scalability-plannerHorizontal/vertical scaling
migration-specialistSystem migration planning

Content & Research (5 agents)

AgentDescription
doc-writerDocumentation generation
api-documenterOpenAPI/Swagger specs
research-analystTechnology research
changelog-authorRelease notes
tutorial-creatorHow-to guides

Building & Development (7 agents)

AgentDescription
contract-spec-authorAPI contract design
integration-specialistSystem integration
modular-builderComponent architecture
module-intent-architectIntent-driven design
performance-optimizerSpeed and efficiency
test-coverageTest analysis and generation
zen-architectProactive architecture guidance

Knowledge & Synthesis (5 agents)

AgentDescription
knowledge-synthesizerCombine information sources
context-builderBuild task context
pattern-recognizerIdentify code patterns
dependency-analyzerDependency graph analysis
codebase-explorerNavigate large codebases

Meta Tools (3 agents)

AgentDescription
amplifier-cli-architectExpert in CLI agent usage
post-task-cleanupCode cleanup after tasks
subagent-architectCoordinate multiple agents

Using Agents

  1. Open the Amplifier panel in the sidebar
  2. Browse or search for an agent
  3. Click to activate
  4. The agent’s expertise is now available in chat

Or mention an agent in chat:

@zen-architect Should we use microservices for this project?

Knowledge Base

Amplifier builds and maintains a knowledge graph of your codebase.

Features

Query Knowledge Ask questions about your code:

  • “Where is authentication handled?”
  • “What patterns does this codebase use?”
  • “How do these modules interact?”

Update Knowledge Refresh the index when code changes significantly:

  • After major refactors
  • When adding new modules
  • After pulling large changes

View Graph Visualize code relationships:

  • Module dependencies
  • Function call graphs
  • Import/export relationships
  • Data flow paths

Show Statistics Track codebase metrics:

  • Files indexed
  • Functions documented
  • Test coverage
  • Complexity scores

Keeping Knowledge Fresh

Knowledge updates automatically as you work, but you can force a refresh:

  1. Open Knowledge Base in Amplifier
  2. Click Update Knowledge
  3. Wait for indexing to complete

CLI Integration

Connect Amplifier to external AI CLIs for expanded capabilities.

Supported CLIs

CLIDescription
Claude CodeAnthropic’s agentic coding assistant
CodexOpenAI’s code-specialized model
GeminiGoogle’s multimodal AI

Switching CLIs

  1. Click the CLI selector at the top of Amplifier
  2. Choose your preferred CLI
  3. The selected CLI handles agent tasks

CLI Status

See availability and version:

  • claude ✓ Available (2.0.72)
  • codex ✓ Available (codex-cli 0.72.0)
  • gemini ✓ Available (0.20.2)

Worktrees

Manage multiple working directories for parallel development.

Creating Worktrees

  1. Open Worktrees in Amplifier
  2. Click Create Worktree
  3. Name it and select a branch
  4. A new isolated workspace is created

Use Cases

  • Feature branches - Work on features in isolation
  • Bug fixes - Hot-fix without disturbing main work
  • Experiments - Try things without affecting your branch
  • Code review - Check out PRs in separate worktrees

Managing Worktrees

  • List All - See all active worktrees
  • Switch Worktree - Change to a different worktree
  • Delete - Remove worktrees you no longer need

Best Practices

Choosing the Right Agent

TaskAgent
Code reviewcode-analyzer, security-architect
Performance issuesperformance-profiler, performance-optimizer
Writing docsdoc-writer, api-documenter
System designzen-architect, api-designer
Bug investigationbug-hunter, error-tracer
Test coveragetest-coverage
Refactoringmodular-builder, pattern-recognizer

Combining Agents

Use multiple agents together:

@zen-architect @security-architect
Review this authentication system design.
Consider both architecture and security.

When to Update Knowledge

  • After cloning a new repo
  • After large merges or rebases
  • When AI seems to have stale context
  • Weekly for active projects