baro-ai
v0.5.1
Published
Autonomous parallel coding - plan and execute with AI
Maintainers
Readme
baro
Background agent runtime orchestrator. Breaks down a goal into stories, builds a dependency DAG, and runs them in parallel — each story gets its own Claude agent.
npm install -g baro-aiHow it works
- You describe a goal
- Claude explores your codebase and plans the work as a dependency graph
- You review and accept the plan
- Stories execute in parallel with live TUI dashboard
- Each story auto-commits and pushes on completion
Usage
# Interactive - opens welcome screen
baro
# Direct - skip to planning
baro "Add authentication with JWT and role-based access control"
# Use OpenAI for planning instead of Claude
baro --planner openai "Add WebSocket support"
# Specify working directory
baro --cwd ~/projects/myapp "Add unit tests"Features
- Parallel execution — independent stories run simultaneously, respecting dependency order
- DAG engine — topological sort with level grouping, cycle detection
- Live TUI — dashboard with story status, live agent logs, DAG view, stats
- Git coordination — mutex-protected commits, auto-push with retry, pull --rebase before each story, conflict detection
- Retry logic — failed stories retry automatically (configurable per story)
- Completion screen — summary overlay with stats when all stories finish
- Claude + OpenAI — Claude as default planner/executor, OpenAI as alternative planner
Requirements
- Claude CLI installed and authenticated
- macOS (arm64/x64) or Linux (x64/arm64)
- Node.js 18+ (only if using
--planner openai)
Architecture
Rust binary distributed via npm. TUI built with ratatui, async execution with tokio, one Claude CLI process per story.
License
MIT
