A playground of Claude Code setups, skills and demos from Anthropic DevRel. Maintained by the Anthropic DevRel team; not accepting external contributions.
Please read before using anything here
- This is not an official Anthropic product. It is a playground: things the Anthropic DevRel team found useful or fun and wanted to share.
- Entries are provided as-is, with no support and no guarantee they still work. The DevRel team maintains the repository, but examples may break as Claude Code, the Claude API and the models change.
- We're not accepting outside contributions yet. Issues and pull requests from outside Anthropic may be closed without review. See CONTRIBUTING.md.
- Nothing here is best-practice canon. Treat it as "here is what we did", not "here is what you should do".
A collection of things the DevRel team is excited to share:
- How we actually set up Claude Code — real hooks,
CLAUDE.mdfiles,settings.jsonexamples and end-to-end personal workflows, with commentary on why. - Demos of what is possible — the result plus the story of how it was made, rather than a template to fork.
- Reusable skills we use ourselves that are not part of the official plugin marketplace.
- Launch demos — demos shown at model launches, with the exact prompt and setup so you can try them yourself.
| Directory | What lives there |
|---|---|
claude-code/hooks/ |
Hook examples, one folder per hook |
claude-code/claude-md/ |
Real CLAUDE.md files with commentary |
claude-code/settings/ |
settings.json examples |
claude-code/workflows/ |
End-to-end personal setups, one folder per person or workflow |
claude-code/mods/ |
Mods: plugins that change what Claude Code does or draw their own UI, one folder per mod |
skills/ |
Reusable skills that are not in the official marketplace |
demos/ |
Art-of-the-possible demos, each with how it was built |
launch-demos/ |
Demos shown at model launches, with the exact prompt and setup |
_template/ |
The entry template and pre-flight checklist every entry follows |
Each entry is self-contained: open its folder and follow the "Run it" section of its README. Nothing is installed or configured at the repository root.
Each entry has its own README.md based on _template/README.md: what it shows, a demo, how it was built (which publicly available model, the prompts, what did not work), how to run it, and its limitations. Before an entry is added it goes through the pre-flight checklist.
| If you want | Go to |
|---|---|
| Agent Skills examples and the skills spec | anthropics/skills |
| Deployable starter apps to clone | anthropics/claude-quickstarts |
| Claude Agent SDK demo apps | anthropics/claude-agent-sdk-demos |
| API notebooks and tutorials | anthropics/claude-cookbooks |
| Installable Claude Code plugins | anthropics/claude-plugins-official |
| Documentation | docs.claude.com and the Claude Code docs |
Entries are staged in a private source repository, which is the source of truth, and exported to the public repository by hand. For each export a maintainer starts from a clean, up-to-date checkout of the private main, reviews the full diff of the export against a checkout of the public repository, and commits and pushes the result themselves. No CI job, bot, webhook or scheduled task ever publishes entries to the public repository; the only automation there is Dependabot, which keeps the pinned GitHub Actions current, and the CLA check. CI only checks; it never publishes.
See SECURITY.md.
Licensed under the Apache License, Version 2.0. See LICENSE. Individual entries may include third-party components under their own licenses; those are noted in the entry.
Claude and Anthropic are trademarks of Anthropic, PBC. Third-party names belong to their owners.