Browse Skills

A step-by-step weekend guide to building a real AI agent from scratch — no coding experience needed. You will learn the core agent loop (goal → tool use → reasoning → result), build a working Python agent using the Claude API, add real tools like web search and file reading, and extend it with your own custom tools. By the end, you will understand how autonomous AI agents actually work by having built one yourself.

A secure sandbox where AI agents can run code safely without touching your real system. When an AI generates and executes code, it happens inside an isolated container — so nothing breaks, nothing leaks. Enterprise-grade infrastructure trusted by leading AI companies.

Add an AI copilot to any React app with a few lines of code. CopilotKit provides ready-made UI components — sidebars, chat panels, text completion — that connect to any AI backend. The standard way to build 'AI-native' features into web products.

Manage multiple AI coding agents — Claude Code, Codex, and others — from one terminal screen. Each agent works in its own isolated environment so they can tackle different features at the same time without conflicting. Switch between them like browser tabs.

Build powerful AI agents with just 100 lines of Python. PocketFlow's minimal design handles complex patterns — parallel processing, document retrieval, human approval steps — without the overhead of larger frameworks. Nothing extra, just what you need.

Give any AI coding agent a long-term memory. Engram stores what Claude Code or Cursor learns across sessions — decisions, patterns, context — and makes it searchable. So next time you open a project, your agent picks up exactly where it left off.

Add structured AI workflows to your project without writing any config files. ZCF automatically manages context windows and breaks down multi-step coding tasks so Claude Code or Codex can handle complex work more reliably.

OpenAI's official toolkit for building multi-agent systems. Define agents with specific roles and tools, then let them hand tasks off to each other automatically. Built-in tracing shows you exactly what each agent did, making it easy to debug and improve.

Build AI workflows that can loop, remember state, and wait for human approval — things a simple chatbot can't do. LangGraph lets you define multi-step agent processes as graphs, so complex tasks like 'research, then write, then review' run reliably.

Connect your local AI coding agent to messaging apps — Slack, Telegram, Discord, LINE, WeChat — without needing a public IP address. Send coding tasks from your phone's chat app and get results back, wherever you are.

A TypeScript framework for building AI agents and workflows with clean, modern code. Made by the Gatsby team, it connects to Claude, OpenAI, and 50+ tools out of the box, with built-in memory management, performance evaluation, and one-command deployment.

Build AI agents in Python the clean, type-safe way. If you like FastAPI's style of structured, validated code, PydanticAI brings the same approach to AI agents. Works with Claude, OpenAI, Gemini, and local models.

Turn your documents — PDFs, notes, databases, emails — into a searchable knowledge base that AI can answer questions about. Ask 'what did we decide in last month's meeting?' or 'summarize this 200-page report' and get accurate, sourced answers.

Pack your entire codebase into a single file that AI can read and understand all at once. Instead of showing Claude one file at a time, Repomix bundles everything — then you paste it in and Claude has full context of your whole project.

Automate Google NotebookLM with code. Upload documents, ask questions, and pull out summaries programmatically — no clicking required. Great for research workflows where you want to process many documents without touching the UI.

Instead of one AI trying to do everything, AutoGen lets you build a team of AI agents that work together — one plans, one codes, one reviews. Microsoft's framework handles the coordination so complex tasks get broken down and finished faster and more reliably.

The most complete library of Claude Code skills, covering GitHub, Slack, Notion, databases, browser automation, and dozens more services. Browse by category, pick what you need, drop it into your Claude skills folder, and Claude Code can now interact with that service directly.

Manage your Supabase database and backend from the terminal. Run a local copy of your database for testing, push changes to production, and auto-generate TypeScript types from your schema — all with simple commands.

The most comprehensive resource hub for Claude Code users. Tutorials, hidden features, community tips, and advanced usage patterns — all in one place. If a technique exists for Claude Code, it's probably documented here.

A community-maintained guide to getting the most out of Claude Code. Covers how to write effective CLAUDE.md files, prompt patterns that actually work, folder organization, and automation hooks — all based on real production experience.

A collection of ready-made skills you can drop into Claude Code to instantly add new abilities. Web scraping, file processing, API calls — each skill adds a new slash command you can run. Copy the folder, restart Claude Code, and the skill is ready to use.

The most widely used toolkit for building AI-powered apps. Connect Claude or other AI models to your documents, databases, APIs, and tools — then chain those connections together into workflows. If you're building a production AI app, chances are LangChain is involved somewhere.

Run Claude Code inside Obsidian as a sidebar chat panel. Built for researchers, it comes with 7 ready-to-use commands: peer review, citation verification, literature search, citation networks, research gap analysis, journal matching, and abstract generation.