Sandbox
Sign in
@MiniCodeMonkey/chief
Harnesses#Claude Code#Codex#Gemini CLI#Cursor#Coding#Product

Task loop harness for Claude Code

Chief is a terminal workflow that breaks a project into tasks and runs an agent through them one by one. It persists progress between runs, so each loop starts fresh while the project keeps moving forward.

474 stars•50 forks•Go•Updated 3mo ago
Who it's for

Builders who want Claude Code or another supported CLI agent to work through a project in task-sized loops.

What it delivers

You can push a big project forward in smaller, reviewable steps without re-explaining the whole job each time.

What it does

Task loop execution

Runs the agent one task at a time until the project plan is complete.

Persistent progress

Saves progress between iterations so the next run can start from where the last one left off.

TUI control

Provides a terminal interface where you can launch Chief and start the loop.

Provider selection

Supports Claude, Codex, Cursor, and Gemini providers through configuration or flags.

One commit per task

Encourages clean git history by committing each completed task separately.

How to get it

  1. 1Run
    brew install minicodemonkey/chief/chief
  2. 2Or via install script
    curl -fsSL https://raw.githubusercontent.com/MiniCodeMonkey/chief/refs/heads/main/install.sh | sh
  3. 3Run
    # Create a new project
chief new

# Launch the TUI and press 's' to start
chief

README

Chief

Chief

Build big projects with Claude. Chief breaks your work into tasks and runs Claude Code in a loop until they're done.

Documentation · Quick Start

Chief TUI

Install

brew install minicodemonkey/chief/chief

Or via install script:

curl -fsSL https://raw.githubusercontent.com/MiniCodeMonkey/chief/refs/heads/main/install.sh | sh

Usage

# Create a new project
chief new

# Launch the TUI and press 's' to start
chief

Chief runs Claude in a Ralph Wiggum loop: each iteration starts with a fresh context window, but progress is persisted between runs. This lets Claude work through large projects without hitting context limits.

How It Works

  1. Describe your project as a series of tasks
  2. Chief runs Claude in a loop, one task at a time
  3. One commit per task — clean git history, easy to review

See the documentation for details.

Requirements

Use Claude by default, or configure another provider in .chief/config.yaml:

agent:
  provider: opencode
  cliPath: /usr/local/bin/opencode   # optional

Or run with chief --agent opencode or set CHIEF_AGENT=opencode.

Supported values for provider are: claude, codex, opencode, cursor, and gemini.

License

MIT

Acknowledgments

Files in the repo

Repository payload•18 top-level entries
  • .github
  • assets
  • cmd
  • docs
  • embed
  • Formula
  • internal
  • ralph
  • .gitignore
  • .goreleaser.yaml
  • CHANGELOG.md
  • CONTRIBUTING.md
  • go.mod
  • go.sum
  • install.sh
  • LICENSE
  • Makefile
  • README.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More harnesses

affaan-m/
ECC
affaan-m/ECCHarnesses

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

258k
ruvnet/rufloHarnesses

🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated

72k
cobusgreyling/loop-engineeringHarnesses

Practical patterns, starters & CLI tools for loop engineering with AI coding agents. Design systems that prompt and orchestrate agents (inspired by Addy Osmani and Boris Cherny). Includes loop-audit, loop-init, loop-cost.

11k