GitHub Repositories Search Scraper - Stars, Topics, Languages

Search GitHub repositories with the full GitHub search syntax and export them as clean rows with stars, forks, open issues, language, topics, license, homepage and dates.

The Actor uses the official GitHub REST API. It works without any token; add your own personal access token to raise the rate limit. A query returning 50 repositories takes about 2 seconds.

What you can do with it

• Find the most starred or most recently active projects on a topic, for example topic:llm language:python.
• Track competitors and alternatives to your project (awesome lists, alternative to X).
• Build datasets of repositories by language, license, size or activity for research.
• Monitor new projects by scheduling the Actor with pushedWithinDays and sortBy: updated.
• Export every public repository of an organization (org:apify) or a user (user:octocat).

Input

Field	What it does
queries	One search per line, in GitHub search syntax: web scraping, topic:llm language:python, org:apify, machine learning stars:>500 pushed:>2026-01-01, license:mit in:readme "openai".
language	Added to every query as language:X.
minStars	Added to every query as stars:>=N.
pushedWithinDays	Keep only repositories with a commit in the last N days.
sortBy	stars (default), forks, updated or best-match.
githubToken	Optional personal access token. Raises the search limit from 10 to 30 requests per minute. No scopes needed for public repositories.
maxItemsPerQuery	Up to 1,000 per query (GitHub's limit).
maxItems	Stop after this many repositories in total.

Popular Python LLM projects active in the last month:

{
  "queries": ["topic:llm"],
  "language": "python",
  "minStars": 1000,
  "pushedWithinDays": 30,
  "sortBy": "updated",
  "maxItemsPerQuery": 200
}

Output

One item per repository:

{
  "id": "570384908",
  "fullName": "huggingface/peft",
  "name": "peft",
  "owner": "huggingface",
  "ownerType": "Organization",
  "ownerUrl": "https://github.com/huggingface",
  "url": "https://github.com/huggingface/peft",
  "description": "PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.",
  "homepage": "https://huggingface.co/docs/peft",
  "language": "Python",
  "topics": ["adapter", "fine-tuning", "llm", "lora", "pytorch", "transformers"],
  "stars": 21791,
  "forks": 2606,
  "watchers": 21791,
  "openIssues": 118,
  "license": "Apache-2.0",
  "isFork": false,
  "isArchived": false,
  "isTemplate": false,
  "defaultBranch": "main",
  "sizeKb": 33811,
  "createdAt": "2022-11-25T03:51:09Z",
  "updatedAt": "2026-10-16T07:55:12Z",
  "pushedAt": "2026-10-15T18:02:44Z",
  "query": "topic:llm",
  "scrapedAt": "2026-10-16T08:28:00.000Z"
}

openIssues includes open pull requests, as in GitHub's own count. watchers equals stars in the search API. license is the SPDX identifier, or the license name when GitHub could not classify it.

Limits and notes

• GitHub returns at most 1,000 results per query. Split broad topics by language, star ranges or dates (created:2025-01-01..2025-06-30) to collect more.
• Rate limits: 10 search requests per minute without a token, 30 with one. Each request returns 100 repositories. The Actor waits for the limit to reset and continues, so large runs take longer rather than failing.
• Only public repositories are returned. The token is never stored or logged.
• README contents, contributors and commit history are not included; use the repository URL with another Actor for those.

Pricing

You pay per repository saved to the dataset, plus a tiny fee per run start.