---
slug: "archi"
source_type: "readme"
source_url: "https://cdn.jsdelivr.net/gh/endlesshoper/archi@main/README.md"
repo: "https://github.com/endlesshoper/archi"
source_file: "README.md"
branch: "main"
---
# Archi

> Project scaffolding and management for vibe coding.

Archi is an [Agent Skill](https://agentskills.io) that helps you set up and manage projects for AI-assisted development. It works with Claude Code, OpenCode, Codex CLI, Cursor, and other skills-compatible agents.

## Installation

```bash
npx skills add EndlessHoper/archi
```

## Usage

Just run `/archi` and the agent will figure out what you need based on context. Or be specific:

| Command | What it does |
|---------|--------------|
| `/archi` | Detect context and choose appropriate action |
| `/archi init` | Initialize new project with docs structure |
| `/archi feature` | Manage features (add/start/complete/update) |
| `/archi status` | Show project overview |
| `/archi docs` | Sync and update documentation |

## Project Structure

When you run `/archi init`, it creates:

```
your-project/
├── SPEC.md           # Product requirements
└── docs/
    ├── features/
    │   ├── todo/     # Backlog
    │   ├── doing/    # In progress
    │   └── done/     # Completed
    ├── research/
    │   ├── spikes/   # Technical experiments
    │   └── references/
    └── bugs/
```

## Feature Tracking

Features are markdown files with YAML frontmatter that move through `todo/` → `doing/` → `done/`:

```yaml
---
title: User Authentication
priority: high
started: 2026-01-23
---
# User Authentication

## Progress
- [x] Basic login
- [ ] OAuth providers
```

The agent detects context from your conversation. Say "I just finished auth" and it'll offer to mark it done.

## Why?

Beginners to AI-assisted coding often struggle with:
- Where to put things
- How to track what they're building
- Keeping the AI informed about the project

Archi gives you an opinionated structure that works well with AI coding tools.

## License

MIT