📈 AI Tool Momentum Index: which AI tools are taking off right now

Stars and likes tell you how big a tool already is. Momentum tells you what is growing this week. This Actor scans GitHub trending repos and Hugging Face trending models and Spaces (or your own list of tools), and ranks every tool with a transparent 0–100 momentum score built from how fast attention is growing: GitHub star velocity, Hugging Face trending and fresh-model downloads, npm and PyPI download growth, and Hacker News coverage.

• ✅ One ranked list across GitHub, Hugging Face, npm and PyPI, AI tools only by default
• ✅ Momentum, not size: stars gained per day, growth rate, downloads of models in their first weeks, week-over-week package growth
• ✅ Hype check: flags models with thousands of likes but almost no downloads, and star spikes with suspiciously few forks
• ✅ Packages found automatically: a GitHub repo is linked to its npm / PyPI package only when the package points back to that repo (no name guessing)
• ✅ Track your own list: repos, models, Spaces, npm and PyPI packages. Schedule it and every run also shows the change since your last run
• ✅ HTML report + clean JSON, ready for Sheets, a newsletter, a dashboard or an AI agent
• ✅ Fair billing: everything found is scored for free, you pay only for the top N tools returned

🎯 Who is it for?

• Developers and tech leads: spot the agent frameworks, inference engines and models your team will be asked about next month.
• VCs, analysts and scouts: open-source traction signals before they show up in funding news.
• Newsletter writers and creators: "the 10 fastest-rising AI tools this week" in one run.
• DevRel and marketing teams: track your project against competitors every day.

🚀 Quick start

1. Leave What to do on Discover what is trending now.
2. Optionally add keywords such as agent, mcp, video or coding.
3. Click Start. Open the HTML report in the Storage tab, or download the dataset as JSON, CSV or Excel.

⬇️ Input examples

This week's trending AI tools on GitHub and Hugging Face:

{ "mode": "trending", "githubPeriod": "weekly", "maxItems": 50 }

Only agent tools, from Python and TypeScript repos:

{ "mode": "trending", "sources": ["github"], "githubLanguages": ["python", "typescript"], "keywords": ["agent"] }

Track your own list (schedule it daily to get the change since the last run):

{ "mode": "track", "tools": ["vllm-project/vllm", "ollama/ollama", "hf:Qwen/Qwen3-8B", "npm:@anthropic-ai/sdk", "pypi:langchain"] }

⬆️ Output (one item per tool)

{
  "rank": 1,
  "kind": "github",
  "id": "alibaba/open-code-review",
  "name": "open-code-review",
  "description": "Secure, fast, efficient, battle-tested at Alibaba's scale...",
  "url": "https://github.com/alibaba/open-code-review",
  "momentumScore": 45,
  "tier": "exploding",
  "signals": ["GitHub: +9,833 stars in 7 days (40,479 total)"],
  "scoreComponents": { "githubVelocity": 35, "githubGrowth": 10, "hfTrending": 0, "hfAdoption": 0, "downloadsGrowth": 0, "hackerNews": 0 },
  "qualityFlags": [],
  "github": { "stars": 40479, "forks": 2912, "starsInPeriod": 9833, "period": "weekly", "starsPerDay": 1404.7, "language": "Go" }
}

Hugging Face items carry huggingFace (likes, trending score, downloads in the last 30 days, task, created date). Packages carry npm or pypi (downloads last week, previous week or monthly average, growth). The key-value store also has OUTPUT (top 10, exploding tools, flagged tools) and REPORT (the HTML report).

Tiers: exploding (45+), rising (30–44), steady (15–29), quiet (under 15).

🧮 How the score works

Component	Max	What it measures
GitHub velocity	35	Stars gained per day (log scale, 1,000/day = max)
GitHub growth	10	Stars gained relative to the repo's size
Hugging Face trending	25	Hub trending score (halved when likes have no downloads behind them)
Hugging Face adoption	20	Downloads in the last 30 days, weighted towards models under 60 days old
Package growth	20	npm week-over-week or PyPI week vs monthly average, only when volume is real
Hacker News	10	Points of stories linking to or naming the tool

Every component is returned with each tool, so you can re-weight it yourself.

💵 Pricing

Free from October 9, 2026. Until then the old pay-per-event prices still apply ($0.02 per tool returned, $0.10 per report). After that date you pay only Apify's normal platform usage for your own runs, typically a few cents.

🤖 Use it from AI agents (MCP)

Add this URL as an MCP server in Claude, Cursor or any MCP client and ask "Which AI tools gained the most momentum this week?":

https://mcp.apify.com?tools=lsso/ai-tool-momentum-index

❓ FAQ

Where does the data come from? GitHub's public trending page and REST API, the Hugging Face Hub API, the npm downloads API, PyPI Stats and the official Hacker News search API (Algolia). No logins, no private data.

Do I need a GitHub token? No. Trending mode works without one. In track mode, a token (read-only public access is enough) lets the Actor measure star velocity for many repos, because GitHub allows only 60 anonymous requests per hour. The token is used only for this run. GitHub lists only the newest 40,000 stargazers of a repo, so for repos with more than 40,000 stars the velocity comes from the change since your previous run (schedule the Actor; the first run shows 0 for them).

Why does a famous model score lower than a new one? The index measures growth, not size. A model with 10 million downloads that is no longer trending scores below a two-week-old model that everyone is trying.

What does the "check" flag mean? A pattern that often means inflated popularity, e.g. a model with thousands of likes and almost no downloads. The tool is still listed; the flag tells you to look twice.

How fresh is it? Every run fetches live data. GitHub trending updates several times a day, Hugging Face trending continuously.

📝 Changelog

• 1.2 (2026-09-25): Becoming free (from October 9, 2026); Hugging Face data now comes only from the official API (the mirror fallback was removed); report links are restricted to http(s).
• 1.1 (2026-09-24): Hugging Face mirror fallback, much lower memory use for popular repos (small npm lookups), 20-second request timeouts, repos with dots in the name (e.g. vercel/next.js) now link to their packages, star losses no longer produce negative scores, clearer message when the charge limit is too low.
• 1.0 (2026-09-24): First release: GitHub trending + Hugging Face models and Spaces, track mode for repos, models, Spaces, npm and PyPI, momentum score with components, hype flags, since-last-run changes, HTML report.