Top 60 Claude Skills, Workflows, and GitHub Repos in 2026
Top 60 Claude Skills, Workflows, and GitHub Repos in 2026
The complete list most people will never find.
Save this before you scroll past. You will thank yourself later. 🔖
The AI tooling landscape in 2026 is unrecognizable. Over 17,000 MCP servers now exist. Anthropic released 17 official skills on GitHub. New agent frameworks ship every single week.
Most of it is noise. Some of it is genuinely useful. A few will permanently change how you work.
I tested 100+ tools over 100+ hours so you do not have to. Here are the 60 that actually matter right now. Organized by category. Tested personally. Honest notes on what each one is actually good for.
Bookmark this. You will come back to it.
Part 1: AI Coding Agents & IDEs 🛠️
The tools that let AI write, review, and manage code on your behalf. The ones that actually work in real workflows, not just demos.
01. Claude Code
Anthropic’s terminal coding agent. Reads files, writes code, runs tests, works directly in your local environment. Powered by Claude Opus 4.6. Scored 80.8% on SWE-bench Verified. The gold standard for AI-assisted development when you want full control.
https://docs.anthropic.com/en/docs/claude-code
02. Cursor
AI-first code editor built on VS Code. Inline completions, chat with your codebase, multi-file editing. 2026 update: run up to 8 parallel agents simultaneously on separate parts of your codebase using git worktrees.
https://www.cursor.com
03. Codex CLI
OpenAI’s terminal coding agent. Takes natural language instructions, reads your codebase, writes and executes code. Strong at multi-step implementation tasks.
https://github.com/openai/codex
04. Windsurf
AI coding IDE by Codeium. Cascade agent for multi-file editing, deep codebase understanding, and flow-state coding. Growing fast in 2026.
https://codeium.com/windsurf
05. Devin
The most autonomous AI software engineer available. Plans, writes code, runs tests, reads docs, fixes bugs, and opens Pull Requests with zero hand-holding. 2026 update: Devin can now orchestrate teams of Devins working in parallel.
https://devin.ai
06. Aider
AI pair programming in your terminal. Works with any LLM. Strong at working with existing codebases. 30,000+ stars.
https://github.com/paul-gauthier/aider
07. Superpowers
20+ battle-tested Claude Code skills. TDD, debugging, plan-to-execute pipelines. 94,000+ stars on GitHub. If you use Claude Code, install this first.
https://github.com/obra/superpowers
Part 2: Agent Frameworks 🤖
Build autonomous systems that think, act, and iterate without you managing every step.
08. OpenClaw
The viral open-source AI agent. Persistent, multi-channel (WhatsApp, Telegram, Discord), writes its own skills. 210,000+ stars and growing. The most accessible entry point for personal AI agents.
https://github.com/openclaw/openclaw
09. LangGraph
Multi-agent orchestration as code. Build agents as graphs with branching logic, human-in-the-loop, and persistent state. 26,000+ stars.
https://github.com/langchain-ai/langgraph
10. CrewAI
Multi-agent framework with roles, goals, and backstories. Each agent has a defined persona and responsibility. Good for team-like automated workflows.
https://github.com/crewAIInc/crewAI
11. AutoGPT
Full autonomous agent platform for long-running tasks. The original agent framework. Matured significantly since its early days.
https://github.com/Significant-Gravitas/AutoGPT
12. Dify
Open-source LLM app builder. Combines workflows, RAG, agents, and model management in one platform. Good for non-developers building AI apps without writing code.
https://github.com/langgenius/dify
13. OWL
Multi-agent cooperation framework. Tops the GAIA benchmark for agent coordination. Cutting edge research turned into usable code.
https://github.com/camel-ai/owl
14. CopilotKit
Embed AI copilots directly into React applications. Ship AI features into your product, not just your personal workflow.
https://github.com/CopilotKit/CopilotKit
15. pydantic-ai
Type-safe agent framework built on Pydantic. For Python developers who want structured, validated agent outputs with full type safety.
https://github.com/pydantic/pydantic-ai
Part 3: MCP Servers & Tool Integration 🔗
MCP (Model Context Protocol) gives AI access to the outside world. Over 17,000 servers now exist. MCP was donated to the Linux Foundation in December 2025, with AWS, Google, Microsoft, Salesforce, and Snowflake as backers. One server works across Claude Code, Cursor, and Windsurf at the same time.
16. Tavily
Search engine built for AI agents. Not blue links. Clean, structured, LLM-ready data. Four tools: search, extract, crawl, map. Connects as a remote MCP in one minute.
https://github.com/tavily-ai/tavily-mcp
17. Context7
Injects up-to-date library documentation into your LLM context. No more hallucinated APIs or deprecated methods. Add “use context7” to your prompt and it pulls current docs. Supports thousands of libraries.
https://github.com/upstash/context7
18. Task Master AI
Your AI project manager. Feed it a PRD and it generates structured tasks with dependencies. Claude executes them one by one. Turns chaotic sessions into organized pipelines.
https://github.com/eyaltoledano/claude-task-master
19. MCP Playwright
Browser automation for LLMs. Control a real browser through natural language. Perfect for testing, scraping, and web interaction.
https://github.com/executeautomation/mcp-playwright
20. Figma MCP
Connect your Figma files directly to Claude Code. Turn design specs into production-ready code with zero manual translation. One of the most talked-about new MCP integrations in 2026.
https://mcp.figma.com/mcp
21. Supabase MCP
Database operations through natural language. Query, insert, and manage your Supabase database directly from Claude. Hosted remotely, zero local setup needed.
https://github.com/supabase-community/supabase-mcp
22. GitHub MCP
Full GitHub access for your AI agents. Read repos, create issues, manage PRs, and trigger actions through natural language. One of the most downloaded MCP servers available.
https://github.com/modelcontextprotocol/servers/tree/main/src/github
23. fastmcp
Build custom MCP servers in minimal Python. The fastest way to create your own tool integrations for Claude or any MCP-compatible model.
https://github.com/jlowin/fastmcp
Part 4: Claude Skills (Top Picks) 🧠
Skills teach Claude specialized workflows. Anthropic now has 17 official skills on GitHub. Over 80,000 community skills exist on SkillsMP. These are the must-install picks from both.
24. PDF Processing (Official)
Read, extract tables, fill forms, merge and split PDFs. The highest-utility skill for knowledge workers.
https://github.com/anthropics/skills/tree/main/skills/pdf
25. Frontend Design (Official)
Build real design systems, bold typography, production-grade UI. Escape the generic AI aesthetic. 277,000+ installs.
https://github.com/anthropics/skills/tree/main/skills/frontend-design
26. Skill Creator (Official)
The meta-skill. Describe any workflow in plain English and get a complete SKILL.md back in five minutes. Build new skills without writing any configuration.
https://github.com/anthropics/skills/tree/main/skills/skill-creator
27. MCP Builder (Official)
Build MCP servers so your LLM can call your own APIs as tools. Four-phase guided workflow built into the skill. New in 2026.
https://github.com/anthropics/skills/tree/main/skills/mcp-builder
28. Claude API Skill (Official)
Multi-language Claude API quickstart. Auto-detects your stack and generates ready-to-run API code across 8 languages. New in 2026.
https://github.com/anthropics/skills/tree/main/skills/claude-api
29. WebApp Testing (Official)
Playwright-based automated UI testing with zero boilerplate. Describe your test in plain English and get automated browser tests back.
https://github.com/anthropics/skills/tree/main/skills/webapp-testing
30. Marketing Skills by Corey Haines
20+ skills covering CRO, copywriting, SEO, email sequences, and growth strategy. Everything a marketing team needs in skill form.
https://github.com/coreyhaines31/marketingskills
31. Claude SEO
Full-site audits, schema validation, keyword analysis. 12 sub-skills covering the complete SEO workflow.
https://github.com/AgriciDaniel/claude-seo
32. Deep Research Skill
8-phase research with auto-continuation. For when you need Claude to go deep on a topic, not just skim the surface.
https://github.com/199-biotechnologies/claude-deep-research-skill
Part 5: Local AI & Model Running 🖥️
Run models on your own hardware. Full privacy, full speed, zero API costs.
33. Ollama
Run open-source LLMs locally with one terminal command. Supports Llama, Mistral, Gemma, and dozens more. The fastest path from zero to local AI.
https://github.com/ollama/ollama
34. Open WebUI
Self-hosted ChatGPT-like interface. Clean, fast, full-featured. Pairs perfectly with Ollama for a fully private AI setup.
https://github.com/open-webui/open-webui
35. LlamaFile
Package an entire LLM as a single executable file. Zero dependencies. Download and run. Absurdly simple.
https://github.com/Mozilla-Ocho/llamafile
36. Unsloth
Fine-tune models 2x faster with 70% less memory. If you need a custom model trained on your own data, start here.
https://github.com/unslothai/unsloth
37. vLLM
High-throughput inference engine. 2 to 4x faster than standard model serving. The standard for production deployment of open-source models.
https://github.com/vllm-project/vllm
Part 6: Workflow & Automation ⚡
Connect AI to your existing tools and business processes.
38. n8n
Open-source workflow automation with 400+ integrations and AI nodes. Self-hostable. 4,000+ community starter templates. The best visual builder for AI-powered automations.
https://github.com/n8n-io/n8n
39. Langflow
Visual drag-and-drop for AI pipelines. 140,000+ stars. Best for building RAG systems, chatbots, and LLM chains as standalone flows.
https://github.com/langflow-ai/langflow
40. Huginn
Self-hosted web agents for monitoring, alerts, and data collection. Privacy-first automation that runs on your own server.
https://github.com/huginn/huginn
41. DSPy
Program (not just prompt) foundation models. Stanford research turned into a production framework. For when prompting alone is not deterministic enough.
https://github.com/stanfordnlp/dspy
42. Temporal
Durable workflow engine for long-running processes. When your automation needs to survive crashes, retries, and timeouts without breaking.
https://github.com/temporalio/temporal
Part 7: Search, Data & RAG 🔍
Get information into and out of AI systems reliably.
43. GPT Researcher
Autonomous research agent that produces compiled reports. Give it a topic, get back a thorough analysis with sources.
https://github.com/assafelovic/gpt-researcher
44. Firecrawl
Turn any website into LLM-ready data. Web scraping designed specifically for AI pipelines.
https://github.com/mendableai/firecrawl
45. Vanna AI
Natural language to SQL. Ask questions in English, get database queries back. For anyone who needs data without writing SQL.
https://github.com/vanna-ai/vanna
46. Instructor
Get structured JSON outputs from any LLM using Pydantic models. Works with OpenAI, Anthropic, Google, and 15+ providers. What production AI engineers actually use.
https://python.useinstructor.com
47. Chroma
Open-source vector database. The simplest way to add semantic search and long-term memory to your AI applications.
https://github.com/chroma-core/chroma
48. dlt
LLM-native data pipelines from 5,000+ sources. Get data from anywhere into your AI workflow with minimal setup.
https://github.com/dlt-hub/dlt
49. ExtractThinker
ORM for document intelligence. Extract structured data from any document type with minimal configuration.
https://github.com/enoch3712/ExtractThinker
Part 8: API & Infrastructure 🏗️
The plumbing that makes everything work in production.
50. FastAPI
The Python web framework for serving AI applications. Exceptional documentation. Pydantic validation built in.
https://github.com/tiangolo/fastapi
51. Portkey Gateway
Route requests to 250+ LLMs through one API. Switch models without changing your code.
https://github.com/Portkey-AI/gateway
52. lmnr
Trace and evaluate agent behavior. See exactly what your agents are doing and whether they are doing it well.
https://github.com/lmnr-ai/lmnr
53. Codebase Memory MCP
Convert your codebase into a persistent knowledge graph. Claude remembers your entire project structure across sessions.
https://github.com/DeusData/codebase-memory-mcp
Part 9: Curated Collections & Learning 📚
Where to find more tools and keep learning.
54. Awesome Claude Skills
The best curated skill list on GitHub. 22,000+ stars. Start here when looking for new skills to install.
https://github.com/travisvn/awesome-claude-skills
55. Anthropic Skills Repo
Official reference implementations from Anthropic. 17 production-ready skills and counting. The gold standard for how skills should be built.
https://github.com/anthropics/skills
56. Awesome Agents
100+ open-source agent tools in one curated list. Broad coverage across every agent category.
https://github.com/kyrolabs/awesome-agents
57. PromptingGuide
Comprehensive prompt engineering reference covering every technique from basics to advanced agent prompting.
https://www.promptingguide.ai
58. Anthropic Prompt Engineering Tutorial
9 chapters of hands-on exercises with Jupyter notebooks. The best structured way to learn prompting from the source.
https://github.com/anthropics/prompt-eng-interactive-tutorial
59. SkillsMP
Marketplace with 80,000+ community skills. The largest catalog for discovering and sharing Claude skills.
https://skillsmp.com
60. Anthropic Official Docs
Covers the API, prompting best practices, tool use, agents, and everything else. Read this before building anything serious.
https://docs.anthropic.com
How to Actually Use This List
Do not try to install all 60 tools at once. That is a recipe for overwhelm and zero results.
Here is the order I recommend:
If you are a developer:
Start with Claude Code (01) + Superpowers (07) + Context7 (17) + GitHub MCP (22). This gives you a powerful AI coding setup with live documentation access and full version control.
If you are a creator or knowledge worker:
Start with OpenClaw (08) + Deep Research Skill (32) + PDF Processing (24) + Frontend Design (25). This gives you an AI assistant with research, document processing, and content creation built in.
If you are building a product:
Start with FastAPI (50) + Instructor (46) + Chroma (47) + LangGraph (09). This gives you the backend framework, structured outputs, memory, and agent orchestration.
If you want to learn first:
Start with the Anthropic Tutorial (58) + PromptingGuide (57) + Anthropic Docs (60). Build the foundation before you stack tools.
Pick one path. Go deep. Add more tools as your needs grow.
TL;DR
Skills = teach AI HOW to do things better
MCP = give AI ACCESS to external tools and data
Repos = the open-source engines powering it all
Combine all three and you have an AI workflow built for real work, not just demos.
💡 Pro Tips
Skills + MCP together are more powerful than either alone
Start with 3 tools max. Go deep before going wide
One MCP server works across Claude Code, Cursor, and Windsurf at the same time
Official Anthropic repos are always the safest and most reliable starting point
Always check the “last commit” date on GitHub before installing anything new
⚠️ The Truth
Most “trending” AI repos are demos, not production-ready tools
90% of tools on GitHub stop getting updates within 60 days
You do NOT need all 60. Pick your category and master it
More tools does not equal better results
The best tool is the one you actually use consistently
🚀 Quick Start
Your GoalStart WithDeveloperClaude Code + Superpowers + Context7 + GitHub MCPCreatorOpenClaw + PDF Processing + Frontend Design + Deep ResearchBuilderFastAPI + Instructor + Chroma + LangGraphLearnerAnthropic Tutorial + PromptingGuide + Anthropic Docs
🎯 Bottom Line
Over 17,000 MCP servers. 80,000+ community skills. 17 official Anthropic skills. New frameworks every week.
You do not need all of it. You need the right stack for YOUR workflow. This list is your shortcut to finding it.
Now go build something.
Comment SKILLS below and I will send you the full categorized bookmark guide. 🔖