artifactize 0.21.1

The AI-native collaboration layer for one-of-a-kind teammates and their agents: split a project into Artifacts with their own evals, and reuse every review while it still holds.
[![CI](https://github.com/lhj6102/artifactize/actions/workflows/ci.yml/badge.svg)](https://github.com/lhj6102/artifactize/actions/workflows/ci.yml)
[![crates.io](https://img.shields.io/crates/v/artifactize.svg)](https://crates.io/crates/artifactize)
[![License](https://img.shields.io/github/license/lhj6102/artifactize)](LICENSE)
[![Download for macOS](https://img.shields.io/badge/Download-macOS-black)](https://github.com/lhj6102/artifactize/releases/latest/download/Artifactize.dmg)
[![Get it on Windows](https://img.shields.io/badge/Get%20it-Windows-0078D4)](https://apps.microsoft.com/detail/9PB6W4LL165D)
[![Download for Linux](https://img.shields.io/badge/Download-Linux%20.deb-E95420)](https://github.com/lhj6102/artifactize/releases/latest/download/Artifactize-amd64.deb)

English · [한국어](assets/readme/README.kr.md)

Homepage: [artifactize.dev](https://artifactize.dev)

# Build together on an Artifact Graph.

Artifactize checks your project piece by piece, from code and documents to images and the
running app, with tests, AI reviews and people's sign-offs. It remembers every result, so
work that hasn't changed is never checked twice, and you can trust what you and your AI
agents changed without watching every step.

Test runners check code and review bots check pull requests. Artifactize puts both, and the
people who sign off, on one Artifact Graph of your whole project: every check knows what it
reads, runs again only when that changes, and keeps going while you step away.

<p align="center">
  <img src="assets/others/roles.en.gif" width="800"
       alt="Animation, collaborate in Artifactize: three roles pass one Artifact Graph between them, made of Artifacts, the pieces of the project, each with its own checks. As Director, three people and two AI agents decide what to build and add an Artifact; as Builder, five AI agents build it; as Evaluator, four AI agents and one person check the Artifacts one by one. The new Artifact is RED, so the Artifact Graph goes back to the Builder to fix, then to the Evaluator again, and it turns GREEN. The Director adds another Artifact and changes one, and the Builder and the Evaluator go back and forth again until GREEN.">
</p>

## Demo

<p align="center">
  <a href="https://github.com/user-attachments/assets/6aafdfbe-e41c-4b90-b435-dab7bfdd28b6">
    <img src="https://raw.githubusercontent.com/lhj6102/artifactize/main/website/demo/media/promo.gif" width="640"
         alt="27-second introduction video, a real terminal recording of Artifactize checking a project and reusing earlier reviews. Select it for the full-quality video.">
  </a>
</p>

## Quick Start

### 1. Install

Pick whichever fits how you work.

**Download the app.** It brings the `artifactize` command with it, so you do not need
Rust's build tools. Your coding agent has nothing new yet: add the plugin below before
your first task.

- **macOS**: [Artifactize.dmg](https://github.com/lhj6102/artifactize/releases/latest/download/Artifactize.dmg),
  one disk image for Apple silicon and Intel. Drag Artifactize to Applications and open it
  once to set up the Terminal command.
- **Windows**: [Artifactize in the Microsoft Store](https://apps.microsoft.com/detail/9PB6W4LL165D)
  (x64), then open a new terminal.
- **Linux**: [Artifactize-amd64.deb](https://github.com/lhj6102/artifactize/releases/latest/download/Artifactize-amd64.deb)
  for x86_64 Ubuntu 22.04, Debian 12 or later: open it with your software installer or run
  `sudo apt install ./Artifactize-amd64.deb`.

Or install from the command line:

| System | Command |
|---|---|
| Linux (x86_64 and aarch64; also WSL 2), desktop app on x86_64 | `curl -fsSL https://artifactize.dev/install.sh \| sh` |
| macOS with [Homebrew](https://brew.sh/) | `brew install lhj6102/tap/artifactize` (add `--cask` for the desktop app too) |
| macOS without Homebrew | `curl -fsSL https://artifactize.dev/install.sh \| sh` |
| Windows, from the Microsoft Store | `winget install 9PB6W4LL165D --source msstore`, then open a new terminal |

The [installation guide](https://artifactize.dev/docs/getting-started/install.html) covers
the PowerShell script, AppImage, Cargo, building from source, updates and uninstalling.

**Add the plugin to your coding agent.** The
[plugin](https://artifactize.dev/docs/agent-plugin/plugin.html) gives Claude Code and Codex
Arti, Artifactize's guide, and Artifactize's tools. It runs the `artifactize` command, so
install that first.

```sh
claude plugin marketplace add lhj6102/artifactize && claude plugin install artifactize@artifactize   # Claude Code
codex plugin marketplace add lhj6102/artifactize && codex plugin add artifactize@artifactize         # Codex
```

**Or let your coding agent set it up.** Paste this into Claude Code or Codex; afterwards
it has the command and the plugin, so you can go straight to your first task:

```text
Install Artifactize on this machine following https://artifactize.dev/docs/getting-started/install.html, then add its plugin to yourself following https://artifactize.dev/docs/agent-plugin/plugin.html.
```

### 2. Try your first task

In your project, ask Arti (in Codex, write `$artifactize:hey-arti`):

```text
/artifactize:hey-arti Explain Artifactize and tell me how we could apply it to this project
```

Arti reads your project, asks what matters to you and proposes Artifacts and checks for it.
It changes your declarations only after you agree, and you get back your project's first
Artifact Graph and a `verify` you can run. To try Artifactize without an AI account, take
the [5-minute quick start](https://artifactize.dev/docs/getting-started/guide.html) with an
example project.

## Highlights

<p align="center">
  <img src="assets/others/graph-example.en.png" width="800"
       alt="Example Artifact Graph: a project's PROBLEM.md, SPEC.md, code/ and docs/ are each declared by an .artf file. A person reviews the problem; AI agents check that the spec solves the problem, that the code's tests check the spec, that the docs explain the code and that the code and docs follow their own rules; the code's tests run as a command.">
</p>

| Feature | What it does |
|---|---|
| Artifact Graph | Each piece of the project, an Artifact, declares its own checks and what it reads, so the graph shows how the pieces connect. |
| Three kinds of checks | A command such as your tests, a review by an AI agent, or a person's sign-off. |
| Reuse | A check whose inputs did not change keeps its earlier result, across versions and teammates' copies. |
| Runs that keep going | Each verification is a Run with its own ID; closing the window does not stop it. |
| Desktop app | Watch Runs live on the graph, open any Artifact and sign off where you are asked to. |

**Who is it for?**

- People who want to build together with others, reviewing efficiently and developing in modules.
- People who want to trust changes made by AI agents without overseeing every step.
- People who want code, documents, images, and running applications to stay consistent as the project grows.

**Our principles**

- We don't restrict how you build. Use your preferred tools and agent harnesses (the apps that run your AI assistants).
- We don't restrict individuality. Create freely within loose guardrails.
- AI doesn't make every decision. Vision, taste, and important decisions about direction remain in human hands.

## The product running

These are real recordings and screenshots of the desktop app, made by scripts
([demos](assets/capture/demos.mts), [screenshots](assets/capture/screenshots.mts)) from
projects they verify for real. The example reviewer "alice" claims a sign-off to show
the review form; nothing is submitted. The text-based app (CLI) also provides Monitor
and Review screens.

**Fix until it passes:** four Runs of an example, eight times faster. Between Runs the
recording script removes what failed, standing in for an AI fixing the code; work that
didn't change is reused at once.

<img src="assets/demos/review-loop.gif" width="1200" alt="Real recording of the Artifactize desktop app over four Runs of one branch: failures turn red, are fixed between Runs, unchanged work is reused, and the fourth Run is satisfied.">

**How this project grew:** Artifactize's own Artifact graph at each release, from
0.7.0 to today.

<img src="assets/demos/graph-growth.gif" width="1200" alt="Real recording of the Artifactize desktop app showing the current Artifact graph of this repository growing release by release from 0.7.0 to the current main.">

**Branches and Runs:** the example's branch with its Run chips and the sign-off waiting
in the left pane, and its Run: core passed, cli's tests failed so docs is blocked, and
release waits for a person's sign-off.

<img src="assets/screenshots/monitor.png" width="1200" alt="Real Artifactize desktop app: the example branch and its waiting sign-off in the left pane, and its Run graph where core passed, cli failed, docs is blocked and release waits for a sign-off.">

**Detail:** cli between the core it uses and the docs that use it, with its checks.

<img src="assets/screenshots/detail.png" width="1200" alt="Real Artifactize desktop Detail of the cli Artifact: its neighbours core and docs, and its lint and tests checks.">

**Review:** the release sign-off, opened from the waiting reviews and claimed by the
example reviewer, after running its notes tool, with the GREEN form open.

<img src="assets/screenshots/review.png" width="1200" alt="Real Artifactize desktop Review: the example's claimed sign-off with its instruction, notes tool, tool log and GREEN form.">

## Learn more

- [Documentation](https://artifactize.dev/docs/): concepts, guides, examples and command reference.
- [Community on Discord](https://discord.gg/VswgKkPkS9): questions, ideas and help.
- [Source and license](LICENSE): open source under Apache License 2.0.