AgentHub ✦

Discover the AI Agent Ecosystem — an open, community-driven directory of AI Agent frameworks, coding agents, browser/computer-use agents, and Agent Skills. Built from scratch with a "data as code" philosophy: the index is generated at build time and the runtime has zero database dependencies.

260 curated projects · 247 repos · 7.4M+ combined GitHub stars · metrics synced daily from the GitHub API

Features

• Home — hero + live ecosystem stats (projects / repos / total stars) + featured projects + category explorer + latest additions
• Browse /browse — debounced keyword search, kind filter (Agent / Skill / Collection), four sort modes, URL state sync
• Categories /c/[category] — 10 categories (Agent frameworks, Prompts & Frontend AI first) with repo-heat overview
• Detail /p/[slug] — stars/forks/maintenance status, one-click install command (npm install for agents, npx skills add for skills), related projects
• Submit /submit — PR-based contribution guide with entry template
• API — GET /api/projects (category/q/kind/sort), GET /api/projects/:slug, GET /api/stats

SEO

• Static prerender — all 230 detail pages + 8 category pages pre-rendered at build time (crawlers get full HTML)
• sitemap.xml — 240+ URLs with lastmod
• Per-page meta — unique title/canonical/description per page, OG & Twitter cards site-wide
• Structured data — WebSite JSON-LD on the home page, SoftwareApplication JSON-LD on every detail page (rich snippets with ratings)
• llms.txt — full-site index for AI search engines (Perplexity, ChatGPT, etc.)

Coverage

Category	Highlights
🤖 Agent Frameworks	LangChain, AutoGen, CrewAI, MetaGPT, OpenClaw, OpenCode, Hermes Agent, Gemini CLI
⚡ Official Skills	anthropics/skills, claude-code, claude-cookbooks, official plugins directory
🛠️ Dev Tools	Addy Osmani's agent-skills, claude-mem, claude-code-router, wshobson/agents
📝 Docs & Writing	academic research skills, best-practice guides
📅 Productivity	marketing, resume, internal comms skills
🎨 Creative	design skills, taste-skill, canvas-design
🔌 Integrations	awesome-mcp-servers, pathway, llmware, quivr, KAG
💬 Prompts	prompts.chat (170k★), Prompt-Engineering-Guide, LangGPT, 中文调教指南, image prompt libraries
⚛️ Frontend AI	screenshot-to-code, OpenUI, Onlook, Playwright MCP, Midscene, magic-mcp
📦 Templates & Collections	awesome lists, marketplaces, starter kits

Plus dedicated ecosystem coverage: memory layers (mem0, Letta, MemOS), deep research agents (Tongyi DeepResearch, gpt-researcher), voice agents (pipecat, LiveKit, TEN), and evaluation/observability (MLflow, Cozeloop).

Quick Start

npm install
npm run data:sync   # fetch GitHub stars/forks -> data/metrics.json (optional; GITHUB_TOKEN recommended)
npm run dev         # http://localhost:3000
npm run build && node .output/server/index.mjs   # production

Architecture

data/projects.ts        # seed data: type-safe entries via defineProjectMeta()
data/categories.ts      # 8 category definitions (order = nav order)
data/metrics.json       # GitHub metrics produced by the sync script (do not edit by hand)
scripts/sync-data.mjs   # GitHub API sync (rate-limit checkpointing, 12h cache, idempotent resume)
scripts/discover.mjs    # batch discovery of candidate repos via GitHub search API
server/api/*            # Nitro API: filter / search / sort / detail / stats
server/routes/*         # sitemap.xml / robots.txt / llms.txt
pages/                  # index / browse / c/[category] / p/[slug] / submit / about
shared/schema.ts        # Project types + defineProjectMeta validation
.github/workflows/      # daily metrics sync (GitHub Actions cron)

GitHub metrics are inlined into the server bundle at build time, so the runtime has zero filesystem or database dependencies — deploys cleanly to Vercel/serverless.

Contributing a Project

Edit data/projects.ts and add an entry:

defineProjectMeta({
  slug: 'my-agent',
  name: 'my-agent',
  description: 'One sentence on what this project does',
  category: 'prompts',          // agents | prompts | official | devtools | docs | productivity | creative | integrations | templates
  kind: 'agent',                // agent | skill | collection
  repo: 'https://github.com/you/my-agent',
  author: 'you',
  tags: ['framework'],
  addedAt: '2026-09-10',
})

Then run npm run data:sync to backfill metrics and open a PR.

Data Sync Strategy

Run npm run data:sync on a schedule (the bundled GitHub Actions cron runs once a day and GITHUB_TOKEN is injected automatically), commit the updated data/metrics.json, and every commit triggers an automatic redeploy on the host.

Deployment (Vercel)

1. Push to GitHub (this repo).
2. On vercel.com/new, import the repo — the Nuxt preset is auto-detected.
3. Add one environment variable: SITE_URL = your production URL (e.g. https://agenthub.vercel.app). This powers canonical URLs, sitemap entries and llms.txt links.
4. Deploy. Every push to main and every daily metrics-sync commit redeploys automatically.

License

MIT