60 Curated Skills, Workflows & Open-Source Projects: The Ultimate Claude Advanced List

marsbitОпубликовано 2026-03-30Обновлено 2026-03-30

Введение

This comprehensive guide curates 60 essential AI tools, frameworks, and resources to help developers, creators, and builders move from using AI to constructing robust AI systems. The list is organized into nine key categories: AI Programming Agents & IDEs (e.g., Claude Code, Cursor), Agent Frameworks (e.g., OpenClaw, LangGraph), MCP Services & Tool Integrations (e.g., Tavily, Context7), Claude Skills (e.g., PDF Processing, Frontend Design), Local AI & Model Runners (e.g., Ollama, Open WebUI), Workflow & Automation (e.g., n8n, Langflow), Search, Data & RAG (e.g., GPT Researcher, Chroma), API & Infrastructure (e.g., FastAPI, Portkey Gateway), and Learning Resources (e.g., Awesome Agents, PromptingGuide). The author emphasizes a structured approach, recommending specific starting toolkits based on the user's role—developer, creator, product builder, or learner—to avoid overload and build practical, production-ready workflows.

Abstract: Suitable for developers, content creators, product builders, and beginners.

Editor's Note: If you've been following AI recently, it's easy to get the illusion: there are more and more tools, but fewer and fewer that you can actually use.

For most people, the difficulty is not whether there are tools, but which tools are worth using and where to start.

This list is based on practical testing and systematically sorts the current AI tool ecosystem, filtering out 60 tools and categorizing them by different levels and usage scenarios: from coding and development tools, to agent frameworks and tool integration, to workflow automation, data processing and infrastructure, and finally extending to learning resources and continuous update paths.

Overall, it outlines a clear path from "Using AI" to "Building AI systems," providing different entry paths for developers, content creators, product builders, and beginners, helping readers find a suitable entry point in the complex tool ecosystem.

In an era of constant tool iteration, such a structured overview is perhaps more meaningful than single-point recommendations. It provides not answers, but a more efficient way to understand and use AI.

The original text is below:

I spent over 100 hours testing various AI tools so you don't have to step on the pitfalls yourself.

Bookmark this first :)

By 2026, the AI tool ecosystem has reached a level of "information overload": new frameworks every week, new agents every day, new projects topping GitHub's trending list every morning.

Most of it is just hype. Some are indeed useful. A very few will truly change how you work.

I've filtered out the noise for you.

The 60 tools below are those I've personally tested and filtered out as currently worth paying attention to—organized by category, along with the scenarios they are truly good at.

Suggest you bookmark it; you'll likely come back to use it repeatedly.

Part 1: AI Programming Agent & IDE

These types of tools allow AI to help you write code, review code, and even manage the entire development process.
The key point: these are ones that can actually run in real workflows, not just stay in demos.

1. Claude Code
A command-line programming agent launched by Anthropic. Can read files, write code, run tests, and operate directly in your local environment.
If you want to maintain full control in AI-assisted development, this is currently the "gold standard".
https://docs.anthropic.com/en/docs/claude-code

2. Cursor
An AI-first code editor built on VS Code. Supports inline completion, conversation with the codebase, and multi-file editing.
Suitable for developers who want to seamlessly integrate AI into existing development workflows.
https://www.cursor.com

3. Codex CLI
A terminal programming agent launched by OpenAI. Give instructions in natural language, it will read the codebase, generate and execute code.
Very strong in multi-step task implementation.
https://github.com/openai/codex

4. Windsurf
An AI programming IDE launched by Codeium. Its Cascade agent supports multi-file editing, deep understanding of the codebase, and an "immersive coding flow".
Growing fast.
https://codeium.com/windsurf

5. Superpowers
A collection of 20+ practical skills built for Claude Code, including TDD, debugging, pipelines from planning to execution, etc.
Over 96k stars on GitHub. If you use Claude Code, it is recommended to install this first.
https://github.com/obra/superpowers

6. Spec Kit (GitHub)
A "Specification-Driven Development" tool: write the specification first, then have the AI generate the code. Forces you to think clearly before starting. Over 50k stars on GitHub.
https://github.com/github/spec-kit

7. Aider
An AI pair programming tool in the terminal, supports any LLM.
Especially strong in handling existing codebases. Over 30k stars on GitHub.
https://github.com/paul-gauthier/aider

Part 2: Agent Frameworks

Used to build automated systems that can "think-act-iterate".

8. OpenClaw
A phenomenal open-source AI agent. Supports long-running operation, multi-channel (WhatsApp / Telegram / Discord), and can even write its own skills.
Over 210k stars on GitHub, currently one of the easiest entry points for personal AI agents.
https://github.com/openclaw/openclaw

9. LangGraph
Uses "graph structures" to orchestrate multiple agents: supports branching logic, human-in-the-loop, and persistent state.
https://github.com/langchain-ai/langgraph

10. CrewAI
A multi-agent collaboration framework where each agent has a role, goal, and "persona".
Suitable for simulating team collaboration processes.
https://github.com/crewAIInc/crewAI

11. AutoGPT
A veteran fully automated agent framework, suitable for long-running tasks.
Much more mature compared to early versions.
https://github.com/Significant-Gravitas/AutoGPT

12. Dify
An open-source LLM application building platform that integrates workflow, RAG, agents, and model management.
Also friendly to non-developers.
https://github.com/langgenius/dify

13. OWL
A multi-agent collaboration framework, leading in GAIA benchmark performance.
Represents cutting-edge research moving towards practical use.
https://github.com/camel-ai/owl

14. CopilotKit
Allows embedding AI copilots directly into React applications.
Not just improving development efficiency, but making AI part of the product.
https://github.com/CopilotKit/CopilotKit

15. pydantic-ai
A type-safe agent framework based on Pydantic.
Suitable for Python developers who want structured, verifiable outputs.
https://github.com/pydantic/pydantic-ai

Part 3: MCP Services & Tool Integration

MCP (Model Context Protocol) allows AI to truly 'connect to the world'. Skills teach it *how* to do things, MCP gives it *the permission to do* them.

16. Tavily
A search engine designed specifically for AI agents, not blue links, but structured data that can be used directly by LLMs.
Offers four tools: search, extract, crawl, and maps; can be integrated with MCP in one minute.
https://github.com/tavily-ai/tavily-mcp

17. Context7
Injects the latest library documentation into the LLM's context.
No more "hallucinated APIs" or outdated methods.
Add a line "use context7" in the prompt to automatically pull the latest docs.
https://github.com/upstash/context7

18. Task Master AI
Your AI project manager. Input a PRD, it breaks it down into tasks with dependencies.
Then executed step-by-step by Claude, turning a chaotic development process into an orderly pipeline.
https://github.com/eyaltoledano/claude-task-master

19. MCP Playwright
Provides browser automation capabilities for LLMs.
Can control a real browser with natural language: testing, scraping, interaction can all be done.
https://github.com/executeautomation/mcp-playwright

20. fastmcp
Quickly build MCP services with minimal Python code.
One of the fastest paths to create custom tool integrations for Claude and other models.
https://github.com/jlowin/fastmcp

21. markdownify-mcp
Converts various formats like PDF, images, audio, etc., into Markdown.
Allows any document to enter the AI workflow.
https://github.com/zcaceres/markdownify-mcp

22. MCPHub
Manages multiple MCP services via HTTP.
A single panel to uniformly manage all tool connections.
https://github.com/samanhappy/mcphub

Part 4: Claude Skills (Curated)

Skills can inject "professional workflow capabilities" into Claude. The community currently has 80k+ skills; the ones below are truly worth installing.

23. PDF Processing (Official)
Supports reading PDFs, extracting tables, filling forms, merging and splitting files.
For knowledge workers, one of the most practical skills.
https://github.com/anthropics/skills/tree/main/skills/pdf

24. Frontend Design (Official)
Used to build truly usable design systems, including bold typography and production-ready UI.
Can break away from the common "AI-generated" design style. Over 277k+ installations.
https://github.com/anthropics/skills/tree/main/skills/frontend-design

25. Skill Creator (Official)
A "meta-skill". Describe your workflow in natural language, generate a complete SKILL.md within 5 minutes.
No need to write configuration to create new skills.
https://github.com/anthropics/skills/tree/main/skills/skill-creator

26. Marketing Skills (Corey Haines)
Covers 20+ skills including CRO, copywriting, SEO, email automation, growth strategies, etc.
Equivalent to "modularizing" the capabilities of an entire marketing team.
https://github.com/coreyhaines31/marketingskills

27. Claude SEO
Provides full-site audits, structured data validation, keyword analysis, and other functions.
12 sub-skills in total, covering the complete SEO workflow.
https://github.com/AgriciDaniel/claude-seo

28. Obsidian Skills
Developed by the Obsidian CEO. Supports automatic tagging, automatic linking, and deep integration with the knowledge base (vault).
If you use Obsidian, this is basically a must-install.
https://github.com/kepano/obsidian-skills

29. Context Optimization
Used to reduce token costs and improve KV cache efficiency.
Can significantly reduce API usage costs. GitHub 13,900+ stars.
https://github.com/muratcankoylan/agent-skills-for-context-engineering

30. Deep Research Skill
8-stage research process + automatic continuation mechanism.
Suitable for scenarios where Claude needs to do "deep research" rather than shallow summaries.
https://github.com/199-biotechnologies/claude-deep-research-skill

Part 5: Local AI & Model Running

Run models on local devices: higher privacy, faster response, no API cost.

31. Ollama
Run open-source large models locally with one command. Supports Llama, Mistral, Gemma, etc.
The fastest path from zero to local AI.
https://github.com/ollama/ollama

32. Open WebUI
A self-hosted ChatGPT-like interface. Simple, fast, fully functional.
Used with Ollama, can build a completely private AI system.
https://github.com/open-webui/open-webui

33. LlamaFile
Packages the entire large model into a single executable file.
No dependencies, download and run, extremely simple.
https://github.com/Mozilla-Ocho/llamafile

34. Unsloth
Increases model fine-tuning speed by 2x, while reducing memory usage by 70%.
If you need to train your own model, this is a great starting point.
https://github.com/unslothai/unsloth

35. vLLM
High-throughput inference engine, performance is 2–4x faster than traditional solutions.
Currently one of the mainstream choices for open-source model production deployment.
https://github.com/vllm-project/vllm

Part 6: Workflow & Automation

Integrate AI into your existing tools and processes.

36. n8n
Open-source workflow automation tool, supports 400+ integrations and AI nodes, can be self-hosted.
Currently one of the strongest visual AI automation builders.
https://github.com/n8n-io/n8n

37. Langflow
Build agent workflows by dragging and dropping. Over 140k stars on GitHub.
No need to write code to build complex agent pipelines.
https://github.com/langflow-ai/langflow

38. Huginn
Self-hosted web agents for monitoring, alerts, and data collection.
Emphasizes privacy-first, all automation runs on your own server.
https://极速赛车开奖结果记录

39. DSPy
Uses "programs" rather than "prompts" to drive models.
A framework originating from Stanford research, suitable for scenarios requiring higher stability.
https://github.com/stanfordnlp/dspy

40. Temporal
A "persistent workflow engine" for long-running tasks.
When your automation processes need to handle crashes, retries, timeouts, etc., this is the standard solution.<极速赛车开奖结果记录>

41. GPT Researcher
Automated research agent that can generate structured reports.
Input a topic, output a complete analysis with source citations.
https://github.com/assafelovic/gpt-researcher

42. Firecrawl
Converts any website into a data format suitable for LLMs.
A crawler tool designed specifically for AI pipelines.
https://github.com/mendableai/firecrawl

43. Vanna AI
Natural language to SQL.
Ask questions in English, directly generates database query statements.
Suitable for those who don't want to write SQL but need to use data.
https://github.com/vanna-ai/vanna

44. Instructor
Uses Pydantic models to make any LLM output structured JSON.
Supports 15+ providers including OpenAI, Anthropic, Google, etc.
The practical choice for many production-level AI engineers.
https://python.useinstructor.com

45. Chroma
Open-source vector database.
One of the simplest ways to add semantic search and long-term memory to AI applications.
https://github.com/chroma-core/chroma

46. dlt
Data pipeline tool for LLMs, can connect to 5000+ data sources.
Helps you import various data into AI workflows.
https://github.com/d极速赛车开奖结果记录

47. ExtractThinker
An "ORM" for document intelligence.
Can extract structured data from any type of document.
https://github.com/enoch3712/ExtractThinker

Part 8: API & Infrastructure

The "underlying pipelines" that make everything run in a production environment.

48. FastAPI
Python web framework for deploying AI applications.
Extremely well-documented, with built-in Pydantic data validation.
https://github.com/tiangolo/fastapi

49. Portkey Gateway
Access 250+ LLMs through one API.
Switch models without changing code.
https://github.com/Portkey-AI/gateway

50. OmniRoute
API proxy supporting 44+ AI providers.
Provides load balancing, failover, and cost optimization.
https://github.com/diegosouzapw/OmniRoute

51. lmnr
Used to track and evaluate agent behavior.
Can clearly see what the agent is doing and measure whether its performance meets standards.
https://github.com/lmnr-ai/lmnr

极速赛车开奖结果记录
Turns your codebase into a "persistent knowledge graph".
Allows Claude to remember the entire project structure across sessions.
https://github.com/DeusData/codebase-memory-mcp

Part 9: Curated Collections & Learning Resources

Where to continuously obtain information and constantly iterate cognition.

53. Awesome Claude Skills
Curated collection of skills, over 22k stars on GitHub.
The首选 entry point for finding new skills.
https://github.com/travisvn/awesome-claude-skills

54. Anthropic Skills Repo
Anthropic's official skills repository.
Also the current "standard paradigm" for skill building.
https://github.com/anthropics/skills

55. Awesome Agents
A curated list summarizing 100+ open-source agent tools.
https://github.com/kyrolabs/awesome-agents

56. PromptingGuide
A complete Prompt Engineering guide covering from basics to advanced.
https://www.promptingguide.ai

57. Anthropic Prompt Engineering Tutorial
Includes 9 chapters + Jupyter Notebook hands-on exercises.
One of the best paths to systematically learn Prompting.
https://github.com/anthropics/prompt-eng-interactive-tutorial

58. SkillsMP
A marketplace platform with 80k+ community skills.
Currently the largest Claude skills directory.
https://skillsmp.com

59. MAGI//ARCHIVE
Daily updates of the latest AI project repositories.
Used to track cutting-edge progress.
https://tom-doerr.github.io/repo_posts/

60. Anthropic Official Documentation
Covers all core content including API, Prompt, tool calls, agents, etc.
If you are serious about building AI products,建议 reading this from cover to cover.
https://docs.anthropic.com

How to Actually Use This List

Don't try to install all 60 of these tools at once. That will only lead to information overload and waste your time.

I recommend using it like this:

If you are a developer: Start with Claude Code (01) + Superpowers (05) + Context7 (17) + Tavily (16).
This combination can help you set up a powerful AI programming environment with search capabilities and documentation support.

If you are a creator / knowledge worker: Start with OpenClaw (08) + Obsidian Skills (28) + PDF Processing (23) + Frontend Design (24). This combination gives you an AI assistant with file management, document processing, and content creation capabilities.

If you are building a product: Start with FastAPI (48) + Instructor (44) + Chroma (45) + LangGraph (09).
This combination covers backend framework, structured output, memory system, and agent orchestration capabilities, enough to support a production-level AI application.

If you just want to learn: Start with the Anthropic tutorial (57) + PromptingGuide (56) + Anthropic official documentation (60).
Lay a solid foundation first, then add tools later.

Choose a path and start深入 using it.

Expand your tools slowly as needs arise.

TL;DR

Skills = Teach AI "how to do better"
MCP = Give AI "the ability to access external tools and data"
Repos = Provide all the open-source infrastructure

Combine these three, and what you get is no longer a "demo that looks cool," but a truly productive AI workflow.

That's it. 60 tools. Now, go build something.

Связанные с этим вопросы

QWhat are the main categories of AI tools covered in the article, and how many tools are listed in total?

AThe article covers 9 main categories: AI Programming Agent & IDE, Agent Frameworks, MCP Services & Tool Integration, Claude Skills (Curated), Local AI & Model Running, Workflow & Automation, Search, Data & RAG, API & Infrastructure, and Curated Collections & Learning Resources. It lists 60 tools in total.

QWhich tool is recommended as the 'gold standard' for developers who want to maintain full control in AI-assisted development?

AClaude Code is recommended as the 'gold standard' for developers who want to maintain full control in AI-assisted development, as it operates directly in the local environment.

QWhat is the purpose of MCP (Model Context Protocol) services and tool integrations mentioned in the article?

AMCP (Model Context Protocol) services and tool integrations allow AI to 'connect to the real world' by giving it the permissions and ability to access and use external tools and data, beyond just knowing how to perform tasks.

QWhat is a recommended starting tool combination for a creator or knowledge worker according to the article?

AFor a creator or knowledge worker, the recommended starting combination is OpenClaw (a versatile AI agent), Obsidian Skills (for knowledge management), PDF Processing (for document handling), and Frontend Design (for content creation).

QWhere does the article suggest developers who are building a production-level AI application should start in terms of tool selection?

AFor developers building a production-level AI application, the article suggests starting with FastAPI (backend framework), Instructor (for structured output), Chroma (vector database for memory), and LangGraph (for agent orchestration).

Похожее

Not a Price Hike, but a Supply Cut? Oil Prices Have Crossed the Tipping Point

The global oil market has passed a critical point, shifting the focus from price increases to potential physical supply shortages. The core issue is a time mismatch: even if the Strait of Hormuz reopens, shipping disruptions have already caused significant delays, which will continue to deplete onshore crude inventories for weeks. Refinery behavior acts as an amplifier. Reduced runs in Asia and Europe don’t reflect weaker demand but instead shrink product inventories, raising fuel prices and refining margins, which in turn encourages higher runs—creating a self-reinforcing cycle. If the Strait remains closed beyond April, traditional pricing models may fail. The market could face an unprecedented physical shortfall of 11-13 million barrels per day—roughly four times historical disruption levels. In such a scenario, price becomes an inadequate balancing tool. The only way to rebalance the market would be policy-driven demand destruction, similar to COVID-era lockdowns. Current prices around $95/barrel are insufficient to balance the market. Key signals to watch include inventory levels, policy announcements, and the pace of involuntary demand contraction. Geopolitically, the situation appears likely to worsen before improving, with little room for compromise between the US and Iran.

marsbit1 ч. назад

Not a Price Hike, but a Supply Cut? Oil Prices Have Crossed the Tipping Point

marsbit1 ч. назад

Amazon Invests Additional $25 Billion in Anthropic, AI Infrastructure 'Arms Race' Escalates

Amazon announces an additional investment of up to $25 billion in Anthropic, with $5 billion delivered immediately and the remaining contingent on performance milestones. This follows a recent $50 billion investment in OpenAI, highlighting Amazon's strategy of backing leading AI labs. The deal includes a commitment from Anthropic to spend over $100 billion on AWS infrastructure over the next decade, securing up to 5 gigawatts of computing power to address growing demand and capacity constraints. Anthropic’s annualized revenue has surpassed $30 billion, but the company faces infrastructure strain due to rapid user growth. The investment will support scaling Claude’s capabilities using Amazon’s custom Trainium and Graviton chips. The move deepens integration between Anthropic and AWS, allowing Claude to be accessed natively within AWS services. Over 100,000 organizations already use Claude via Amazon Bedrock. This investment is part of a broader AI infrastructure race, with Amazon planning around $200 billion in capital expenditures this year, largely focused on expanding AI compute capacity.

marsbit1 ч. назад

Amazon Invests Additional $25 Billion in Anthropic, AI Infrastructure 'Arms Race' Escalates

marsbit1 ч. назад

A New CEO Who Has Worked Exclusively with Hardware for 25 Years Takes Over Apple, Valued at 4 Trillion

Apple, the world's most valuable tech firm, has appointed John Ternus as its new CEO, effective September 1, replacing Tim Cook who transitions to executive chairman. Ternus, a 51-year-old mechanical engineer with 25-year tenure at Apple, previously served as senior vice president of hardware engineering. Known for his low public profile and lack of social media presence, he played key roles in developing products like the iPad, AirPods, and led Apple’s transition from Intel to in-house silicon chips. Ternus takes over a company valued at $4 trillion, with Cook having multiplied Apple’s revenue and built a $100 billion services business. However, he inherits significant challenges: Apple’s AI efforts, including the delayed Siri revamp powered by Google’s Gemini, remain unproven, and the company is preparing to launch its first foldable iPhone amid supply constraints. While Ternus’s hardware expertise positions him well for product-driven innovation, his lack of software/AI experience raises questions about Apple’s competitiveness in the AI era. The company’s structure has been adjusted to support him, with Cook remaining in an advisory role. Ternus’s engineer-led approach—emphasizing humility and collaboration—may prove vital in navigating Apple’s next chapter.

marsbit1 ч. назад

A New CEO Who Has Worked Exclusively with Hardware for 25 Years Takes Over Apple, Valued at 4 Trillion

marsbit1 ч. назад

Atkins' First Year at the Helm of the SEC: A Comprehensive Shift in Crypto Regulation

Paul Atkins marked his one-year anniversary as Chair of the U.S. Securities and Exchange Commission (SEC) on April 21, 2025, overseeing a significant shift in the agency’s approach to cryptocurrency regulation. Under his leadership, the SEC dropped multiple enforcement actions against crypto firms, approved several crypto-linked ETFs, and issued guidance clarifying that most cryptocurrencies are not considered securities under federal law. The SEC also signed a memorandum with the CFTC to improve regulatory coordination. These actions reversed the aggressive enforcement stance of his predecessor, Gary Gensler, and aligned with Trump administration promises to support the crypto industry. However, Atkins has faced criticism from Democratic lawmakers, including Senator Elizabeth Warren, who raised concerns over potential conflicts of interest, particularly regarding dropped cases linked to Trump-affiliated companies. While regulatory clarity has improved, the SEC still awaits congressional action to formally define its jurisdiction over digital assets.

marsbit1 ч. назад

Atkins' First Year at the Helm of the SEC: A Comprehensive Shift in Crypto Regulation

marsbit1 ч. назад

a16z: 5 Ways Blockchain Can Help AI Agent Infrastructure

Blockchain technology provides critical infrastructure for AI agents by addressing five key challenges: 1) Non-human identity: AI agents lack standardized, portable identity systems. Blockchain enables verifiable, cross-platform agent identities (like "Know Your Agent" frameworks) through cryptographic credentials and on-chain registries. 2) AI governance: When AI systems execute decisions, blockchain ensures transparency and prevents centralized control by recording actions on-chain and enabling auditable execution logs. 3) Payments: Stablecoins and crypto payments (e.g., x402, MPP) serve as default settlement layers for agent-to-agent commerce, enabling frictionless, programmable transactions for "headless" AI-native businesses. 4) Trust and verification: As AI scales, blockchain provides cryptographic proof of origin and auditable histories, making verification—not intelligence—the scarce resource. 5) User control: Crypto-native tools (e.g., delegation toolkits, intent-based architectures) allow users to set boundaries and maintain oversight over autonomous agents, minimizing blind trust. Together, blockchain and AI can create an economic infrastructure built on transparency, accountability, and user sovereignty.

marsbit2 ч. назад

a16z: 5 Ways Blockchain Can Help AI Agent Infrastructure

marsbit2 ч. назад

Торговля

Спот
Фьючерсы
活动图片

Популярные категорииright-arrow

Популярные тегиright-arrow