Official training from Anthropic, OpenAI, Google, and Hugging Face alongside independent courses on agent frameworks. Learn tool use, retrieval, MCP servers, and agentic workflows that actually run in production.
Sorted cheapest first. Click a column header to re-sort. Developers and builders.
| Course | Level | Duration | Price | Certificate | |
|---|---|---|---|---|---|
HumanLayer (Dex Horthy) · GitHub |
intermediate | 12 short chapters plus a bonus factor, a few hours of reading | Free | No | Visit |
Google · Kaggle |
intermediate | 5 days (self-paced) | Free | No | Visit |
Hugging Face |
intermediate | ~15-20 hours (5 units plus bonus units) | Free | Yes | Visit |
Microsoft · GitHub |
beginner | 18 lessons, self-paced | Free | No | Visit |
Geoffrey Huntley · GitHub |
intermediate | 6-step workshop, a few hours | Free | No | Visit |
Anthropic · Anthropic Academy |
intermediate | Self-paced (multi-module with final assessment) | Free | Yes | Visit |
Anthropic · Anthropic Academy |
intermediate | Self-paced (4 modules) | Free | Yes | Visit |
Anthropic · Anthropic Academy |
beginner | Self-paced (~12 lessons) | Free | Yes | Visit |
Anthropic · Anthropic Academy |
intermediate | 85 lectures, ~8 hours of video, 10 quizzes | Free | Yes | Visit |
OpenAI · OpenAI Academy |
intermediate | ~110 minutes (lessons plus guided lab) | Free | Yes | Visit |
Microsoft · Microsoft Learn |
intermediate | 9 modules, ~10 hours | Free | No | Visit |
Nir Diamant (DiamantAI) · GitHub |
intermediate | 64 notebook tutorials, self-paced | Free | No | Visit |
LangChain · LangChain Academy |
intermediate | 46 lessons (~30 minutes of video plus hands-on labs) | Free | No | Visit |
LangChain · LangChain Academy |
intermediate | ~6 hours of video (55 lessons) | Free | No | Visit |
Anthropic · Anthropic Academy |
intermediate | Self-paced | Free | Yes | Visit |
shareAI-lab · GitHub |
intermediate | 17 progressive lessons, self-paced | Free | No | Visit |
Hugging Face and Anthropic · Hugging Face |
intermediate | 4 units plus bonus; ~3-4 hours per week, about a week per unit | Free | Yes | Visit |
Microsoft · GitHub |
beginner | 13 modules, self-paced | Free | No | Visit |
Anthropic · Anthropic Academy |
advanced | 15 lectures, ~1 hour of video, 2 quizzes | Free | Yes | Visit |
Scrimba (Bob Ziroll) · Scrimba |
intermediate | ~2 hours (117 minutes) | Paid $24/mo |
No | Visit |
DeepLearning.AI, Google Cloud and IBM Research · DeepLearning.AI |
intermediate | ~1.5 hours (14 lessons, 8 code examples) | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI and Anthropic · DeepLearning.AI |
beginner | ~2.3 hours (10 video lessons) | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI |
intermediate | ~10 hours (5 modules, 31 video lessons) | Free to audit Free to audit |
Yes | Visit |
Microsoft · edX |
intermediate | 2 weeks, ~7 hours per week | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI, LangChain and Tavily · DeepLearning.AI |
intermediate | ~1 hour 42 minutes | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI and LangChain · DeepLearning.AI |
beginner | ~1 hour 53 minutes | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI and LangChain · DeepLearning.AI |
intermediate | ~1.3 hours (8 video lessons, 4 code examples) | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI and Anthropic · DeepLearning.AI |
intermediate | ~1 hour 58 minutes | Free to audit Free to audit |
Yes | Visit |
DeepLearning.AI and CrewAI · DeepLearning.AI |
beginner | ~3 hours (18 lessons, 7 code examples) | Free to audit Free to audit |
Yes | Visit |
Udacity |
intermediate | ~53 hours (4 courses, 4 projects) | Paid $249/mo |
Yes | Visit |
No payment required. Several of these are complete university courses.
An influential essay series, modelled on the 12-Factor App, arguing for owning your prompts, context window and control flow, treating tools as structured outputs, and keeping agents small, stateless and pausable. It suits engineers moving agents from demo to production who want design principles rather than another framework tutorial.
Google and Kaggle's intensive pairs a daily whitepaper with hands-on codelabs on foundation models, embeddings, agents, domain-specific LLMs, and MLOps using Gemini and the Agent Development Kit. For developers who want a dense, free bootcamp-style sprint through Google's GenAI stack.
Covers agent fundamentals (tools, reasoning, actions), then builds agents with smolagents, LangGraph, and LlamaIndex, finishing with a benchmark challenge and leaderboard. For Python developers who want framework-agnostic agent skills and a free certificate.
Microsoft's 18-lesson open course on agent design patterns, tool use, planning, multi-agent systems, agentic RAG, memory, trust and security, and protocols like MCP and A2A. For developers who want a free, vendor-documented survey of agent concepts with runnable samples.
A six-step workshop that grows a basic Claude chat program into a working coding agent by adding file reading, directory listing, shell execution, file editing and code search, explaining the agent event loop along the way. It suits developers comfortable with Go who want to demystify coding agents by writing a minimal one themselves.
Anthropic's comprehensive developer course spanning API setup, prompt evaluation, tool use, RAG, extended thinking, prompt caching, MCP, and agent workflows like routing and parallelization. The go-to course for Python developers who want to build production apps on Claude.
Shows developers how to run longer, less supervised Claude Code sessions: plan mode, CLAUDE.md, skills, permission modes, hooks, headless automation, GitHub Actions, and team plugins. For engineers already using Claude Code who want to move to team-wide, verifiable workflows.
A ground-up introduction to the Claude Platform using the TypeScript SDK: first requests, model selection, agent loops, tool use, built-in tools, skills, MCP, and Managed Agents. For developers who have never used an LLM API and want the fastest route to a working agent.
A long technical course on building with Claude through Amazon Bedrock: the API via boto3, prompt evaluation and engineering, tool use, RAG with embeddings and reranking, extended thinking, prompt caching, MCP, and agent design with Claude Code and computer use. For backend, ML and DevOps engineers on AWS who want a free end-to-end path to Bedrock-hosted Claude agents.
Design multi-agent applications with clear roles, controlled tool access and reliable handoffs, then test the workflow against defined outcome and safety requirements in a guided lab. For developers with API experience who want OpenAI's own method for building agentic systems on its platform.
Builds agents with Microsoft Foundry Agent Service and Microsoft Agent Framework: custom tools, MCP tools, knowledge grounding with Foundry IQ, publishing to Teams and Microsoft 365 Copilot, workflows, multi-agent orchestration and A2A discovery. For developers and AI engineers building agents on Azure who want Microsoft's free, code-first path (course AI-3026).
A large, community-favourite collection of 64 runnable notebooks, from a beginner conversational agent to multi-agent research, business and QA systems, using LangGraph, PydanticAI, CrewAI, AutoGen, MCP and plain SDK code. It suits Python developers who learn best by reading and modifying working agent code across many frameworks.
Builds long-running agents with Deep Agents, LangChain's open-source agent harness: models, tools, MCP, human-in-the-loop, filesystem backends and sandboxes, summarization, skills and memory, subagent delegation, and a sales-assistant capstone in Python or TypeScript. For developers building complex multi-step agents who want LangChain's current recommended approach.
LangChain's official LangGraph course: graphs, routers and agents, state and memory, streaming and human-in-the-loop breakpoints, parallelization and sub-graphs, long-term memory with the LangGraph Store, and deployment. For Python developers who want precise control over agent workflows, taught by the team that builds the framework.
Teaches the MCP architecture and how to build servers with the Python SDK, exposing tools, resources, and prompts, then testing them with the MCP Inspector. For Python developers who want to connect Claude (or any MCP client) to their own data and services the standard way.
A 0-to-1 course that rebuilds a Claude Code-style coding agent one mechanism per lesson (tool use, permissions, hooks, planning, subagents, context compaction, memory, task systems, scheduling, agent teams, MCP plugins and a goal loop) with runnable Python and diagrams. It suits developers who use coding agents daily and want to understand, or build, the harness around the model.
Built by Hugging Face in partnership with Anthropic, it covers MCP architecture and core concepts, then an end-to-end use case and a deployed use case, plus bonus units, with tools such as Gradio. For developers who want a free, hands-on MCP course with free certificates to complement Anthropic's own MCP courses.
A 13-module curriculum on MCP fundamentals, security, building servers and clients, VS Code integration, testing, deployment, and advanced topics like OAuth2 and scaling, with samples in C#, Java, JavaScript, Python, Rust, and TypeScript. For developers in any language who want a thorough free grounding in MCP.
Goes beyond MCP basics into sampling, log and progress notifications, roots-based file access control, JSON message types, and the STDIO and StreamableHTTP transports, including state handling for production servers. For developers who finished Introduction to Model Context Protocol and now need to ship MCP servers and clients to production.
Watch everything free; pay only if you want graded assignments and a credential.
Builds a healthcare multi-agent system from agents written in different frameworks, wraps each in an A2A server, builds A2A clients, and orchestrates them in sequential and hierarchical workflows using Google ADK, LangGraph, BeeAI and Microsoft Agent Framework. For developers who need agents from different teams or vendors to discover and talk to each other.
Anthropic's Elie Schoppik explains the SKILL.md format, progressive disclosure, and how skills compare with tools, MCP and subagents, then builds custom skills for Claude.ai, Claude Code, the Claude API and a research agent on the Claude Agent SDK. For builders who want one reusable way to give agents specialized know-how across Anthropic's products.
Andrew Ng's course on building agentic workflows in Python around four design patterns (reflection, tool use, planning and multi-agent collaboration), plus evaluation, error analysis and production deployment. DeepLearning.AI's most complete framework-agnostic agent course, for Python developers who want to design agents that work reliably.
Microsoft's code-first edX course on building AI agents with Semantic Kernel and Azure AI services: function calling and custom tools, RAG with Azure AI Search, multi-agent orchestration, and deploying and monitoring production agents. For developers who want a structured course with an optional paid certificate rather than a docs-style learning path.
Build an agent from scratch, then rebuild it in LangGraph with agentic search, persistence, streaming, and human-in-the-loop controls, ending with an essay-writing agent. For developers with intermediate Python who want controllable, graph-based agents.
Harrison Chase and Andrew Ng teach the LangChain building blocks: prompts and parsers, memory, chains, question answering over documents, evaluation, and agents. A quick on-ramp for Python developers who want to build their first LLM application.
Introduces four loops around an AI agent (the agent loop, verification loop, event-driven loop and hill-climbing loop) and adds each in turn to a requirements-review agent built with LangChain Deep Agents. For developers who already have a working agent and want systematic ways to make it more reliable over time.
Elie Schoppik of Anthropic walks through MCP's client-server architecture, building a FastMCP server with tools, resources, and prompts, wiring an MCP client into a chatbot, and connecting Claude Desktop. For developers who want to make their apps MCP-compatible in an afternoon.
CrewAI founder João Moura teaches how to design agents and organize them into crews with roles, memory, tools, guardrails and sequential, parallel or hierarchical cooperation, applied to six business processes such as resume tailoring and event planning. For Python users who want a fast start with the CrewAI framework for multi-agent automation.
Structured programs with instructors, cohorts, projects or formal credentials.
An interactive Scrimba screencast course in which learners pause the video and edit the instructor's code in the browser while building LLM-powered agents in JavaScript. It suits front-end and JavaScript developers who prefer typing along in the browser over watching long videos, and who already have or want a Scrimba Pro subscription.
A project-based program moving from reasoning and planning prompts to agentic workflow patterns, tool-using agents with memory and agentic RAG, and multi-agent orchestration, with four portfolio projects including a multi-agent trip planner and sales system. For Python developers who want reviewed projects and a structured, paid path into agent engineering.
Transformers, large language models, fine-tuning, and building with foundation models.
Write prompts that reliably get the results you want from any model.
Deployment, monitoring, data pipelines, and engineering practices for real AI systems.