Skip to content

blueman82/conductor

Conductor Logo

Conductor

Autonomous Multi-Agent Orchestration for Claude Code

Conductor Demo

Conductor orchestrates multiple Claude Code agents to execute implementation plans in parallel waves. It parses Markdown/YAML plans, calculates dependencies, spawns agents, and manages quality control reviews with automatic retries.

Features

Run in Smithery

  • Wave-Based Execution — Parallel tasks within waves, sequential between waves
  • Quality Control — Automated reviews with GREEN/RED/YELLOW verdicts and retries
  • Adaptive Learning — Learns from history, swaps agents on failures
  • Intelligent Selection — Claude picks best agents for tasks and QC
  • Multi-File Plans — Cross-file dependencies, split large plans
  • Runtime Enforcement — Test commands, criterion verification
  • Pattern Intelligence — STOP protocol for prior art detection, duplicate prevention
  • Mandatory Commits — Agents instructed to commit, conductor verifies via git log
  • Budget & Rate Limits — Intelligent auto-resume, session resume, state persistence
  • Voice Feedback — Optional TTS via local Orpheus server

Quick Start

# Install (macOS Apple Silicon)
curl -L https://github.com/blueman82/conductor/releases/latest/download/conductor-darwin-arm64 -o conductor
chmod +x conductor && sudo mv conductor /usr/local/bin/

# Validate and run a plan
conductor validate plan.md
conductor run plan.md --verbose

See releases for other platforms.

Usage

conductor validate plan.md          # Check syntax and dependencies
conductor run plan.md               # Execute the plan
conductor run plan.md --dry-run     # Preview without executing
conductor run plan.md --task 3      # Run only task 3
conductor run plan.md --skip-completed --retry-failed  # Resume

Key flags: --max-concurrency N, --timeout 10h, --verbose, --log-dir PATH

Plan Format

Markdown:

## Task 1: Setup
**File(s)**: setup.go
**Depends on**: None

Create project structure.

## Task 2: Implementation
**File(s)**: main.go
**Depends on**: Task 1
**Success Criteria**:
- Compiles without errors
- Tests pass

Implement main logic.

YAML:

plan:
  tasks:
    - id: 1
      name: Setup
      files: [setup.go]
      description: Create project structure.
    - id: 2
      name: Implementation
      files: [main.go]
      depends_on: [1]
      success_criteria: ["Compiles", "Tests pass"]
      description: Implement main logic.

Configuration

Create .conductor/config.yaml:

max_concurrency: 3
timeout: 10h
quality_control:
  enabled: true
  retry_on_red: 2
  agents:
    mode: intelligent
learning:
  enabled: true
  swap_during_retries: true
guard:
  enabled: true
  mode: warn  # block | warn | adaptive
  probability_threshold: 0.7
pattern:
  enabled: true
  mode: warn  # block | warn | suggest
  enable_stop: true
  require_justification: true
tts:
  enabled: false
  base_url: "http://localhost:5005"
budget:
  enabled: true
  auto_resume: true           # Wait for rate limit reset
  max_wait_duration: 6h       # Save state if wait exceeds this

Documentation

Status

Production-ready — 86%+ test coverage, 500+ tests.

License

MIT

About

No description, website, or topics provided.

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Packages

No packages published

Contributors 3

  •  
  •  
  •