Here’s today’s look at what developers are starring on GitHub, taken from the daily trending page and the Python, JavaScript, TypeScript and PHP trending lists on the evening of 9 November 2026 (IST). Star counts are exactly as GitHub showed them at that time. They change quickly, so expect higher numbers when you visit.

AI agents dominate the list again, but there’s also a physics simulator, a voice model and a useful new PHP SDK.

How we pick

GitHub’s trending lists rank repositories by how many stars they gained recently, not by total stars. That makes them a good signal for what is new or newly popular. We skip repositories that are mainly about collecting personal data on individuals, tools built to avoid content safety controls, and anything we can’t describe accurately from the project’s own page. We also skip anything already covered in an earlier roundup on this blog.

Trending repositories

1. anthropics/skills

Language: Python · Stars: 190,472 (1,052 today)

The public repository for Agent Skills: ready-made skill folders (a SKILL.md plus optional scripts and resources) that agents load on demand. Why devs care: Agent Skills was donated to the Linux Foundation’s Agentic AI Foundation in October, so skills are becoming a cross-vendor format. This repo is the best place to see real-world examples before writing your own.

2. meta/astryx

Language: TypeScript · Stars: 18,982 (3,324 today)

Described by the project as “an open source design system that’s fully customizable and agent ready.” Why devs care: it was today’s fastest climber overall. A design system built to be used by AI coding agents as well as people is worth watching if you build front-ends with agent help.

3. vercel-labs/agent-browser

Language: Rust · Stars: 48,247 (914 today)

A browser automation CLI for AI agents. Why devs care: agents that can drive a real browser can test web apps, fill forms and collect data. A CLI tool is easy to plug into almost any agent setup or CI pipeline.

4. google-deepmind/mujoco

Language: C++ · Stars: 16,791 (579 today)

MuJoCo (Multi-Joint dynamics with Contact) is a general-purpose physics simulator. Why devs care: it’s a standard tool in robotics and reinforcement-learning research. If you’re experimenting with RL environments or robot control, this is the engine many papers and benchmarks use.

5. opensandbox-group/OpenSandbox

Language: Python · Stars: 16,804 (159 today)

A “secure, fast, and extensible sandbox runtime for AI agents.” Why devs care: running agent-generated code safely is one of the hardest parts of building agents. OpenSandbox was also one of the four projects that joined the Agentic AI Foundation in October, alongside Agent Skills.

6. OpenBMB/VoxCPM

Language: Python · Stars: 42,063 (423 today)

VoxCPM2 is a tokenizer-free text-to-speech model for multilingual speech generation, voice design and voice cloning. Why devs care: open TTS models keep getting better, and a multilingual one is useful for apps that serve Indian and global audiences. As with any cloning tool, use it only with consent.

7. OpenCut-app/OpenCut

Language: TypeScript · Stars: 97,905 (201 today)

“The open-source CapCut alternative.” Why devs care: a large, active TypeScript codebase for a real video editor is a great project to study or contribute to, especially if you’re interested in media handling in the browser.

8. modelcontextprotocol/php-sdk

Language: PHP · Stars: 1,666 (2 today)

The official PHP SDK for building Model Context Protocol (MCP) servers and clients, maintained in collaboration with The PHP Foundation. Why devs care: it was at the top of today’s PHP trending list. If you run Laravel or Symfony apps, this is the official way to expose your app’s data and actions to AI assistants through MCP.

GitHub changelog: latest updates

There were no new entries on the GitHub Changelog in the last 24 hours at the time of writing. The newest posts are from Friday, 6 November. Here are the recent ones most worth knowing:

  • @GitHub in Slack and Teams is generally available (6 Nov). You can investigate bugs, create issues, implement changes and open pull requests from a chat conversation. @GitHub now works out the repository from your message.
  • Copilot code review: API management (6 Nov). Copilot code review settings can now be managed through the REST and GraphQL APIs, from enterprise policy down to individual repositories. The review overview comment is also more compact.
  • New pull request Overview page in public preview (5 Nov). An opt-in page that puts merge status and next available actions at the top in a compact summary, so you don’t have to scroll to the merge box on long PRs.
  • Dependabot supports Python 3.15 (3 Nov). You can upgrade projects to Python 3.15 and keep receiving Dependabot updates.
  • Heads-up: pull_request_target blocked by default from 2 December (2 Nov). GitHub Actions will block pull_request_target events by default in all public repositories unless an Actions event policy explicitly allows them. Private and internal repositories aren’t affected. If your open-source project relies on this trigger, review your workflows now. The rule runs in evaluate mode first, so you can see what would be blocked.

How to explore trending repos well

Stars aren’t a quality guarantee. Before adding any of these projects to real work, check four things: the licence, how recently it was committed to, how many people maintain it, and whether issues get responses. For agent tools in particular, read what permissions they need and where they send data. A tool that controls a browser or runs code deserves the same review as any other dependency with access to your systems.

If you want to follow trends for your own stack, GitHub’s trending page has a language filter and daily, weekly and monthly views. Checking the weekly view for your main language is a low-effort way to discover useful libraries early.

Takeaway

Today’s trending list shows where developer attention is: tools that make AI agents more capable (skills, browser control) and safer (sandboxes). On the platform side, the most urgent item is the pull_request_target change, so if you maintain a public repo, check your workflows before 2 December.