The only Claude Code orchestrator that auto-secures, containerizes, and parallelizes your entire feature.
Overwhelm features with coordinated AI workers. Plan, design, rush, merge — all in parallel.
pip install zerg-ai
Key Features
Secure by Default
Auto-fetched OWASP rules. Language-specific security. Zero manual config.
Run Anywhere
Docker containers, subprocess, or task mode. Git worktrees for isolation.
Massively Parallel
Coordinated workers with dependency-aware scheduling. 5+ workers building simultaneously.
Token Smart
Context engineering: command splitting, scoped budgets. 30-50% token savings per worker.
Why ZERG?
Manual CLAUDE.md for every project
Auto-detects stack, fetches security rules, generates configs
One agent, sequential execution
5+ workers building in parallel with dependency-aware scheduling
Context rot across long sessions
Spec-driven workers — stateless, restartable, no conversation history
Token waste on repeated context
Context engineering: command splitting + scoped budgets = 30-50% savings
Quick Start
From zero to parallel workers in four commands.
# Install ZERG
pip install zerg-ai
# Plan your feature
/zerg:plan user-auth
# Design architecture and task graph
/zerg:design
# Launch the swarm
/zerg:rush --workers=5
Install — Get ZERG from PyPI with a single pip command.
Plan — Interactively capture requirements for your feature.
Design — Auto-generate architecture and parallel task graph.
Rush — Launch 5 parallel workers to build your feature.
How It Works
Step 1
Plan
Capture requirements through interactive questioning
Step 2
Design
Generate architecture and atomic task graph
Step 3
Rush
Launch parallel workers with exclusive file ownership
Step 4
Merge
Quality gates, automated merges, level by level
26 Commands. One Toolkit.
Every command you need to plan, build, test, and ship.
Core Workflow
| Command | Description |
|---|---|
| /zerg:brainstorm | Open-ended feature discovery with competitive research |
| /zerg:plan | Capture requirements through interactive questioning |
| /zerg:design | Generate architecture and task graph for parallel execution |
| /zerg:rush | Launch parallel workers to execute task graph |
Monitoring & Control
| Command | Description |
|---|---|
| /zerg:status | Display real-time execution progress |
| /zerg:logs | Stream, filter, and aggregate worker logs |
| /zerg:merge | Manually trigger level merge operations |
| /zerg:retry | Retry failed or blocked tasks |
| /zerg:stop | Stop workers gracefully or forcefully |
| /zerg:cleanup | Remove ZERG artifacts and free resources |
Quality & Analysis
| Command | Description |
|---|---|
| /zerg:analyze | Static analysis, complexity metrics, and quality assessment |
| /zerg:build | Build orchestration with auto-detection and error recovery |
| /zerg:refactor | Automated code improvement and cleanup |
| /zerg:review | Two-stage code review workflow |
| /zerg:security | Security scanning and rules management |
| /zerg:test | Execute tests with coverage and generation |
Utilities
| Command | Description |
|---|---|
| /zerg:create-command | Scaffold new ZERG commands with Task integration |
| /zerg:debug | Deep diagnostic investigation with error intelligence |
| /zerg:git | Git operations with intelligent commits, PRs, and releases |
| /zerg:plugins | Extend ZERG with quality gates, hooks, and launchers |
| /zerg:worker | Internal worker execution protocol |
Documentation & AI
| Command | Description |
|---|---|
| /zerg:document | Generate documentation for a single component |
| /zerg:estimate | PERT effort estimation with post-execution comparison |
| /zerg:explain | Educational code explanations with four depth layers |
| /zerg:index | Generate complete documentation wiki |
| /zerg:select-tool | Intelligent tool routing across MCP, native tools, and agents |
0
Commands
Complete CLI toolkit
0
Execution Modes
Container, subprocess, task
0%
Token Savings
Context engineering efficiency
Frequently Asked Questions
What is ZERG?
ZERG is a parallel execution system for Claude Code. It coordinates multiple Claude instances to work on features simultaneously, splitting tasks by file ownership to eliminate merge conflicts.
What do I need to get started?
Python 3.10+, Claude Code CLI, and optionally Docker for container mode. Install with pip install zerg-ai and run /zerg:init inside Claude Code.
Is it free?
ZERG is open source under the MIT license. You'll need your own Anthropic API key or Claude Pro/Team subscription for the Claude Code instances.
How does it handle API keys?
For subprocess/task mode, workers inherit your Claude Code session. For container mode, mount ~/.claude (OAuth) or pass ANTHROPIC_API_KEY as an env var.
Do I need Docker?
No. ZERG's default task mode runs entirely within Claude Code using the Task tool. Docker is only needed for container mode isolation.
How does ZERG compare to other tools?
ZERG is purpose-built for Claude Code with spec-driven workers, exclusive file ownership (no merge conflicts), and context engineering for token efficiency. It's not a general CI/CD tool — it's an AI orchestrator.