Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
| Date | Stars |
|---|---|
| 2026-07-31 | 257 |
| 2026-08-03 | 257 |
| 2026-08-13 | 256 |
| 2026-09-17 | 257 |
| 2026-09-20 | 257 |
Today
— stars today
This week
+1 stars this week
This month
+1 stars this month
Momentum
0.0
growth rate 0.39%/day
# Ghost
**Commit intentions, not code.**
Ghost is a CLI that flips the git workflow: instead of committing code, you commit *prompts*. An AI coding agent generates the artifacts; the commit captures both the intent and the output. Your git history becomes a chain of prompts + their results.
Supports **claude**, **gemini**, **codex**, and **opencode** — swap agents per-commit or set a default.
<img width="935" height="657" alt="image" src="https://github.com/user-attachments/assets/cc33324e-7a33-4572-b9ab-be1b3d773c0e" />
## The Idea
Code is ephemeral. Intent is permanent.
Every `ghost commit` answers: *what did I want to happen here?* Not *what bytes changed*. Each commit is reproducible from its prompt — if the code breaks, you have the exact instruction that generated it. The git log reads like a design document, not a diff summary.
```
git log --oneline
a3f2c1b add JWT authentication middleware
7e91d4a create user registration endpoint with email validation
2bc0f88 scaffold Express app with TypeScript and Prettier
```
Each of those is a ghost commit. Behind each message is an AI that turned words into working code, and a session ID that ties the output back to the generation.
## Why Intent-Based Commits?
**Code is the artifact, intent is the source of truth.**
When you read a traditional git log, you see *what* changed. With ghost, you see *prompts* — the human decision that triggered the change. A year from now when LLMs are more amazing you can replay the git log and generate a better version.
**Every commit is reproducible.** The prompt is preserved with some extra attributes about which model and agent was used. You can re-run any commit against a fresh checkout to see what Claude generates from the same instruction.
**The log becomes a design document.** Read `ghost log` top-to-bottom and you'll see the intent behind every architectural decision, not just the code that resulted from it.
**Diffs show what the AI decided; messages show what you asked for.** The two together give you full context: the goal and the implementation, inseparably linked.
## How It Works
```
you: ghost commit -m "add user auth with JWT"
↓
agent generates code → files written to working tree
↓
ghost detects changes → stages new/modified files
↓
git commit with enriched message (prompt + agent + model + session + file list)
```
## Quick Start
```bash
git clone <this-repo>
export PATH="/path/to/ghost/bin:$PATH"
cd your-project
ghost init
ghost commit -m "create a REST API endpoint for user registration"
```
## Commands
| Command | Description |
|---|---|
| `ghost init` | Init git repo (if needed), install hook, create `.ghost/` dir |
| `ghost commit -m "prompt"` | Generate code from prompt, stage changed files, commit |
| `ghost commit --agent gemini -m "prompt"` | Use a specific agent (claude, gemini, codex, opencode) |
| `ghost commit --dry-run -m "prompt"` | Generate code, show what changed, don't commit |
| `ghost log` | Pretty-print ghost commit history (prompt, agent, model, session, files) |
| `GHOST_SKIP=1 ghost commit -m "..."` | Pass-through to plain `git commit` |
### Examples
```bash
# New feature (default agent: claude)
ghost commit -m "add a login page with email/password form and client-side validation"
# Use Gemini
ghost commit --agent gemini -m "refactor the database layer to use connection pooling"
# Use Codex
ghost commit --agent codex -m "fix the race condition in the payment processing queue"
# Use OpenCode
ghost commit --agent opencode -m "add OpenAPI documentation for all endpoints"
# With a specific model
ghost commit --agent claude --model claude-opus-4-6 -m "architect a microservices migration plan as code comments"
# Preview without committing
ghost commit --dry-run -m "add OpenAPI documentation for all endpoints"
# Set default agent via env
GHOST_AGENT=gemini ghost commit -m "scaffold a new service"
# Manual commit (bypass ghost entirely)
GHOST_SKIP=1 ghost commit -m "bExcerpt of 5,909 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
Not classified yet. Classification runs as part of npm run ingest.