Artifacto — Semantic Artifacts for AI Coding Agents

Give your AI agent a map of your codebase — before it starts working.

Artifacto automatically creates and maintains structured JSON "artifacts" for every module in your project. The Cursor/VS Code AI agent reads these artifacts first, understands your code's intent, and avoids breaking your architecture.

The Problem

Every new AI session starts blind. The agent:

• Doesn't know which files are critical
• Doesn't know what patterns are intentional ("don't touch this")
• Has to grep through 50 files to understand one feature
• Makes changes that break things it didn't know existed

The Solution

Artifacto creates artifacts/ in your project:

artifacts/
  _index.md              ← project overview (agent reads this first)
  trajectory.md          ← where the project is heading
  .changes.jsonl         ← log of recent file changes
  modules/
    auth.json            ← what auth.py does, its deps, what NOT to touch
    api_server.json      ← entry points, known risks, changelog
    database.json        ← ...and so on for every module

The agent reads _index.md → finds the right module artifact → goes to the actual code. Instead of grepping 50 files, it reads 3.

Features

✅ Free (all users)

• Auto-initialization — one command creates artifacts for the entire project
• AST-based extraction — fills artifacts from source code without AI (Python, TypeScript, JavaScript, Lua, Go, and more)
• Real-time updates — artifacts update automatically when files change
• Change log — artifacts/.changes.jsonl tracks what changed since last session
• Eager startup refresh — artifacts are fresh before your first message to the agent
• Health Check — diagnoses issues with your setup
• Cursor rules auto-install — agent is forced to read artifacts via .cursor/rules/
• .cursorignore setup — module JSONs excluded from semantic search (no pollution)
• CI/pipeline support — python artifacts/refresh.py for automated workflows

🔒 Pro (coming soon)

• AI Fill — use LLM to enrich artifacts with deeper semantic understanding
• Trajectory analysis — AI-powered project direction from git history
• Team sync — share artifact snapshots across your team

Quick Start

1. Install Artifacto
2. Open your project in Cursor/VS Code
3. Ctrl+Shift+P → Artifacto: Initialize Artifacts
4. Done — artifacts are created and the agent is configured automatically

The agent now reads artifacts/_index.md at the start of every session.

How It Works

Your code
  │
  ▼
Artifacto (AST parser)
  │
  ▼
artifacts/modules/auth.json
{
  "file": "src/auth.py",
  "purpose": "JWT authentication middleware",
  "entry_function": "verify_token(token: str) -> User",
  "dependencies": ["jwt", "database"],
  "dont_touch": "token expiry logic — intentionally strict",
  "known_risks": "rotating keys requires cache flush"
}
  │
  ▼
.cursor/rules/artifacto-core.mdc
"HARD STOP: read artifacts/_index.md before any response"
  │
  ▼
AI Agent with full context

Supported Languages

Status Bar

Commands

For Automated Pipelines / CI

# Refresh artifacts without human interaction
python artifacts/refresh.py

# Force-refresh all modules
python artifacts/refresh.py --all

Requires Python 3.8+ and the Artifacto extension installed.

Requirements

• Cursor or VS Code 1.85+
• Python 3.8+ (for AST extraction — no API keys needed)

Beta

Artifacto is currently in public beta. We're actively improving based on user feedback.

Get Lifetime Pro free: Install Artifacto, use it on a real project, and share your feedback → open a GitHub issue with what worked, what didn't, and what you'd want next. First 200 beta testers who do this get Lifetime Pro when it launches.

Privacy

Artifacto runs entirely locally. No code is sent anywhere. The AST extractor is pure Python with no network calls. Artifacts stay on your machine.

Built for developers who take AI-assisted coding seriously.