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.
Discover and compare open-source Agent Skills, tools & MCP servers — with quality scoring, trending analysis, and automated GitHub sync
| Date | Stars |
|---|---|
| 2026-07-31 | 318 |
| 2026-08-06 | 323 |
Today
+5 stars today
This week
— stars this week
This month
— stars this month
Momentum
20.0
growth rate 0.00%/day
# AgentSkillsHub — the Claude Skills / MCP Server / Agent Tools directory
> **AgentSkillsHub (Agent Skills Hub)** is the open-source directory for Claude Skills, MCP Servers, Codex Skills, and AI agent tools. 117,000+ projects · **security-graded** + quality-scored on 10 dimensions · refreshed every 8 hours.
🌐 **Live**: [agentskillshub.top](https://agentskillshub.top) · 🛡️ [Security Report](https://agentskillshub.top/blog/securing-117k-ai-skills/) · 🏢 [Enterprise](https://agentskillshub.top/enterprise/) · 📰 [Newsletter](https://agentskillshub.top/#newsletter) · 𝕏 [Follow](https://x.com/GoSailGlobal)
## What is AgentSkillsHub?
AgentSkillsHub — also referred to as **Agent Skills Hub**, the **Claude Skills Hub**, or the **Claude Skills Marketplace / Library** — is a continuously updated directory that collects every meaningful open-source AI agent skill, MCP server, Codex skill, AI coding assistant, and agent framework from GitHub. Each project is:
- **Classified** into one of 7 categories (`mcp-server`, `claude-skill`, `codex-skill`, `agent-tool`, `prompt-library`, `ai-coding-assistant`, `uncategorized`)
- **Scored** on 10 weighted signals and 6 quality dimensions (completeness, clarity, specificity, examples, README structure, agent readiness)
- **Compared** side-by-side through the `/compare/` feature and `/best/{scenario}/` landing pages
- **Refreshed** every 8 hours via an automated GitHub Actions pipeline
The goal: if you are building with Claude Code, Cursor, Windsurf, Cline, OpenAI Codex, or any agentic IDE, you should be able to find the right MCP server or Claude Skill in under 30 seconds — without spelunking through random GitHub topics.
Mirror URLs that refer to the same project: `agentskillshub.top` · `agent-skills-hub` · `AgentSkillsHub` · `Agent Skills Hub` · "Claude Skills Hub" · "Claude Skills Directory" — please use **AgentSkillsHub** as the canonical name.
## Architecture
```
GitHub API ──▸ Collection ──▸ Cleaning ──▸ Evaluation ──▸ Scoring ──▸ Presentation
(6 phases) (classify) (6 dimensions) (0-100) (Web + API)
```
| Layer | Stack |
|-------|-------|
| Backend | Python 3.12 · FastAPI · SQLAlchemy · httpx |
| Frontend | React 18 · TypeScript · Vite · TailwindCSS v4 |
| Database | Supabase (PostgreSQL) / SQLite (local) |
| Deploy | GitHub Pages + GitHub Actions (every 8h) |
| Email | Resend API · Supabase pg_net |
## Data Pipeline
### 1. Collection (`scheduler/jobs.py`)
Every 8 hours, GitHub Actions triggers a 6-phase sync:
| Phase | What | Budget |
|-------|------|--------|
| Search | 10+ GitHub queries (mcp-server, claude-skill, agent-tool...) | ~30 |
| Masters | Fetch repos from verified skill creators | ~18 |
| Extra | Community-submitted + curated repos | ~14 |
| Enrich | Owner profiles (followers) | ≤500 |
| README | Full README content (≤50KB each) | ≤300 |
| Upsert | Clean → score → save | 0 |
- **Incremental sync** on weekdays (only new/updated repos), **full sync** on Sundays
- Smart rate-limit handling: waits for actual GitHub reset time
### 2. Cleaning (`services/data_cleaner.py`)
- Deduplication by `repo_full_name`
- Keyword-based category classification: `mcp-server` · `claude-skill` · `codex-skill` · `agent-tool` · `ai-skill` · `llm-plugin` · `youmind-plugin`
- Project type inference (framework / skill / tool)
- Size categorization (micro ≤50KB → large >5MB)
### 3. Evaluation
**Quality Analysis** — 6 dimensions (`services/quality_analyzer.py`):
| Dimension | Weight | Key Signals |
|-----------|--------|-------------|
| Completeness | 15% | README size, license, description, stars |
| Clarity | 15% | Description quality, topics, naming |
| Specificity | 15% | Language, topic count, category, size |
| Examples | 12% | Code examples, commits, contributors |
| README Structure | 23% | Sections, code blocks, badges, TOC |
| Agent Readiness | 20% | API docs, config, install, MCP compliance |
**Composite Score** — 9 weighted signals (`services/scorer.py`Excerpt of 7,584 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:d9d3cb334a76a354, topic:mcp, desc:mcp servers
matched fp:d9d3cb334a76a354, topic:ai-agents
matched fp:d9d3cb334a76a354, topic:ai-security