AI skills that matter

Copilot suggests lines, whole functions, and code blocks as you type, right inside VS Code, JetBrains, and Neovim. It trains on GitHub public repositories and understands your project context. Chat mode lets you ask about code, generate tests, and explain errors without leaving the editor.

A VS Code fork that puts AI at the center of every operation. Beyond autocomplete, you can select code and request natural-language changes, ask questions about a whole file or the entire codebase. Composer mode lets you write complete features from a single prompt, modifying multiple files simultaneously.

Anthropic's official CLI that turns Claude into an agent that can read, write, execute commands, and commit to your local repository. Unlike editor plugins, it operates on the entire codebase without context window limits. Ideal for large refactors, complex debugging, and features spanning dozens of files.

Codeium's editor introduces the "Cascade" concept: the AI doesn't wait for your commands but anticipates next steps and proposes chained edits. Great for those wanting a more automated experience than Cursor. Remote mode can connect to servers and containers, useful for distributed dev environments.

Open-source CLI tool that connects to any model (GPT-5.5, Claude, Llama) and works directly in your terminal and Git repo. Every change is proposed as a diff and applied only after confirmation. Supports whole-codebase maps to keep context on large projects. Free and completely local.

LangChain is the starting point for anyone building LLM-powered apps. It provides abstractions for prompt chains, vector database integration, memory management, and external tool calls. Massive community and thousands of ready integrations. Great starting point; migrate to LangGraph for more complex cases.

LangChain extension that models agent workflows as directed graphs. Each node is a function, each edge a conditional transition. Enables loops, branching, and persistent state across calls — essential for agents that need to think multiple steps, retry, and maintain memory across sessions. Used in production by major AI companies.

Framework for orchestrating multiple AI agents each with a precise role (researcher, writer, reviewer) that collaborate to complete complex tasks. Define the team in Python, assign goals, and watch agents divide the work. Great for content pipelines, multi-step data analysis, and business process automation.

Microsoft open-source framework for building multi-agent conversations where models talk to each other to solve problems. Supports agents with code execution, web browsing, and API calls. AutoGen Studio offers a visual interface for building agent teams without writing Python code.

Pydantic framework for building AI agents with structured outputs automatically validated via Python types. Eliminates malformed JSON issues: define a Pydantic schema and the agent always returns conforming data. Supports all major models (OpenAI, Anthropic, Gemini) with the same interface.

Anthropic's flagship model. It leads coding benchmarks (tops SWE-bench Pro) and writes the most natural prose around, holding very long manuscripts together thanks to up to 128k output tokens. Huge context, great at following complex instructions without hallucinating, with a "computer use" API that lets agents control real browsers and desktops. Available via API and on Claude.ai.

OpenAI's flagship: it leads the general intelligence index and agentic terminal benchmarks, with a 1 million-token context window. Natively multimodal (text, images, audio in one architecture), great for screenshot analysis, charts, and voice assistants. Available in ChatGPT (free tier included) and via API. The 'mini' variant costs a fraction for simple tasks.

Google's flagship: it leads scientific-reasoning benchmarks (GPQA Diamond) with a context window of roughly 1 million tokens. Multimodal and natively integrated into Google Workspace. For speed and low cost there's the Gemini 3.5 Flash variant. Free up to generous limits with the Gemini API.

Meta's open-weight family, the first to use a Mixture-of-Experts (MoE) architecture, and the most widely deployed open ecosystem in the world. Llama 4 Scout has the largest context window among open models (up to 10M tokens); Llama 4 Maverick beats previous flagship models on coding, reasoning, and images. Download it and run it locally without sending data to anyone, or use Groq/Together.ai for ultra-fast cloud inference.

Mistral AI's (France) most capable model: an open-weight multimodal Mixture-of-Experts with 41B active parameters out of 675B total, near the top of the open leaderboards. Excels at multi-step reasoning, code, and European languages — often more natural than American competitors in Italian and French. The Codestral version specializes in coding. Self-host it (open weights) or use the Mistral API at competitive prices.

AI-powered search engine that answers in natural language while always citing original sources — eliminates hallucination issues because every claim is traceable. Perfect for quick research on recent news, academic papers, and verifiable facts. Pro version accesses GPT-5.5 and Claude for more elaborate responses.

Perplexity alternative optimized for technical and programming questions. Automatically searches Stack Overflow, GitHub Issues, official docs, and technical blogs, then synthesizes the answer with code examples. Has its own model (Phind-70B) trained on code that outperforms GPT-4 on coding benchmarks.

Web search API designed to be used by AI agents, not humans. Unlike Google, Exa understands semantic queries ("find articles similar to this") and returns results with full page content, not just links. Ideal for building agents that need to gather web information in a structured way.

AI search engine with an ecosystem of built-in "apps" (coder, writer, research) accessible from the same search box. Strong privacy focus: use without account and opt out of tracking. Research mode does deep multi-site searches and produces structured reports, similar to ChatGPT Deep Research.

Open-source Zapier/Make alternative you can install on your own server and keep completely private. Over 400 ready integrations, native AI nodes for GPT/Claude, and the ability to write JavaScript in nodes for custom logic. Cloud plan is free up to 5 active workflows. The choice for technical users who want full control.

Visual automation platform where you build workflows by connecting icon "modules" with drag & drop. Easier than n8n for non-technical users, with over 1000 supported apps. Has built-in AI modules for GPT, Claude, and Gemini. Great for automating emails, CRM, social media, and notifications. Free plan with 1000 ops/month.

Platform for creating AI agents without writing code, optimized for business use cases: customer support, lead qualification, document analysis. Each "agent" is configurable with tools, knowledge bases, and custom prompts. Integrates with CRM, Slack, email, and can be triggered via API or webhook.

The veteran of no-code automation has integrated AI in two ways: "AI Actions" lets you use GPT to transform data mid-Zap, while "Zapier Agents" creates conversational bots connected to all your apps. If you already use Zapier for automation, adding AI costs little and requires zero code.

Image generation model with the most recognizable style and most active community. Operates via Discord (or beta web interface) with a text prompt system. Version 6 produces near-photographic images and better understands text in images. /describe analyzes an existing image and returns prompts to recreate it.

The leading platform for cloning and synthesizing human voices. With 30 seconds of audio you can create a clone of your voice to use for podcasts, videos, or voice assistants. Supports 30+ languages with natural accents. Has an API developers use to give AI agents a voice. Free plan for 10k characters/month.

Generates complete musical tracks (vocals, instruments, mix) from a simple text description. "A 90s rock ballad about loneliness" produces a sung and played track in seconds. Version 4 has significantly improved vocal quality and stylistic consistency. Free plan with 50 credits per day.

AI video studio with three main capabilities: Gen-3 Alpha generates video from text or image (up to 10 seconds), "Motion Brush" adds movement to specific parts of a static image, and AI editing tools remove objects, change backgrounds, and apply artistic styles. The choice of AI video professionals.