Skip to content

cs3b/ace

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8,255 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ACE - Agentic Coding Environment

Workflow harness for coding agents — and the developers who work with them.

ACE Logo

Ruby License: MIT

Quick Start | Vision | Architecture | Tools Reference

Works with: Claude Code, Codex CLI, OpenCode, Gemini CLI, pi-agent, and more.

The Problem

Coding agents can write code, but they still lack the workflow layer around the code:

  • Your tasks live in a SaaS dashboard your agent can't see
  • Context is scattered across files the agent doesn't know to load
  • Review feedback is verbal and untracked — nothing saved, nothing actionable
  • Multi-step workflows require you to babysit every transition

ACE gives agent harnesses like Claude Code and Codex CLI the workflow layer they're missing — task pipelines, context bundles, review presets, and skills — as standard CLI commands both you and your agents run the same way.

What ACE Does

Tasks and ideas as files your agent can read

  • Capture ideas and turn them into task specs with subtasks.
  • Track progress in markdown files that live in your repo.
  • Let any agent read the same task state you do.
ace-idea create "Add retry logic to webhook delivery" --tags reliability
ace-task create "Implement webhook retry" --priority high

ace-task | ace-idea

One command from task to shipped PR

  • Provision an isolated worktree.
  • Open a tmux window for focused execution.
  • Kick off a multi-step assignment that runs implement, test, review, and ship.
ace-overseer work-on --task 8r3

ace-overseer orchestrates ace-assign, ace-git-worktree, and ace-tmux

Multi-model code review with tracked feedback

  • Run the same diff through multiple LLM providers.
  • Compare review output side by side.
  • Track suggestions through a verified, pending, and resolved lifecycle.
ace-review --preset code-deep --pr 123

ace-review

Context loading on demand

  • Load project context, workflow instructions, and guides through wfi://, guide://, and tmpl://.
  • Bundle context for agents or load it directly yourself.
ace-bundle wfi://task/work

ace-bundle | ace-handbook | ace-search

Install

Start small with smart commits, or install the full orchestration stack:

gem install ace-git-commit    # smart commits, zero config
gem install ace-overseer      # full orchestrator stack

Ruby 3.2+ required. Each ace-* gem installs independently, so you can start with one tool and add more as needed.

Principles

  • CLI-first, agent-agnostic — every tool is a standard CLI command; if it can run bash, it can use ACE
  • Everything in git — tasks, workflows, configs, and review sessions are markdown and YAML files in your repo
  • Transparent and inspectable — dry-run support, readable configs, saved sessions; when something breaks, you can trace it
  • Customizable cascade — each package ships defaults; override at user (~/.ace/) or project (.ace/) level without forking

The Toolkit

40+ packages total, including shared libraries and agent platform integrations. See Tools Reference for the complete inventory.

Agent Platform Support

ACE skills project natively to multiple agent platforms. Install an integration package and workflows, guides, and skills appear in your agent's expected directory structure.


Quick Start | Vision | Architecture | Tools Reference

About

ACE — modular CLI toolkit for building agentic development environments (ADE). Same tools for developers and AI agents — installable Ruby gems with consistent, predictable interfaces.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages