A LinkedIn post rounds up 20 AI repos worth knowing
A LinkedIn post built a list of 20 GitHub projects ("repos," meaning a folder of code stored on GitHub) tied to AI, ranked by how many people have starred, meaning bookmarked, each one. It is the kind of practical list you have flagged before, useful for spotting something worth cloning.
The first group is AI coding agents: tools that write and edit code for you on their own. OpenClaw leads the whole list with 278,000 stars, more than anything else on it. Behind it: Opencode (118,000), Claude Code (75,000), Superpowers (73,000), and Codex (63,000).
Next is agent engineering and tools, the plumbing used to build and run agents rather than the agents themselves: Firecrawl (89,000 stars), Context7 (48,000), Scrapling (25,000), Agent Browser (19,000), and Symphony (8,900).
The AI infrastructure group, the software layer that actually runs AI models, has Open WebUI (126,000), llama.cpp (97,000), Daytona (63,000), and Zeroclaw (24,000).
The last two groups are for learning: Hermes Agent (200,000), Awesome LLM Apps (100,000), AI Agents for Beginners (53,000), Prompt Engineering (32,000), and Hello Agents (25,000) teach the basics, while System Prompts of AI Tools (129,000) is a single curated collection of the actual instructions other AI products run on.
For scale, the post notes GitHub hosts 450 million repos in total, so even the smallest name here, at 8,900 stars, sits well above the ordinary project.