Free AI courses from the companies that build the models

Last checked November 9, 2026. Confirm prices and free access on each provider's site.

Which one first?

New to AI at work: AI Fluency: Framework and foundations, then Claude 101.
Moving from a non-coding role such as BA, QA or PM: AI Fluency for builders, then the agents guide on this site.
Comfortable with code: Building with the Claude API or OpenAI's API pathway, then an evals course.

Pick one of these first. They teach how to work with AI well, not how to code.

AI Fluency: Framework and foundations (Anthropic): Introduces the 4D framework (Delegation, Description, Discernment, Diligence). 14 lessons, about 4 hours.
AI capabilities and limitations (Anthropic): A mental model of what language models can and cannot do. About 3.5 hours.
Claude 101 (Anthropic): Everyday use, from prompting to projects and connected tools. About 2.5 hours.
AI Foundations pathway (OpenAI): OpenAI Academy lists AI Foundations, Applied AI Foundations, and Agents and Workflows in its Foundations pathway.
Get an introduction to generative AI (Microsoft): A short Microsoft Learn module.

For the part of an AI Solutions Engineer role that involves building. Read our guide to agents first.

AI Fluency for builders (Anthropic): Owning a project from problem to shipped solution. 9 lessons, about 3 hours.
Introduction to agent skills (Anthropic): Building, configuring and sharing reusable skills. About 1 hour.
Introduction to subagents (Anthropic): Splitting a complex task across parallel subagents. About 45 minutes.
Design and Build Agentic Systems (OpenAI): Part of the API pathway on OpenAI Academy.
Agentic AI (DeepLearning.AI): Listed as intermediate level, about 10 hours.
Develop your first agent with Microsoft Foundry (Microsoft): A hands-on Microsoft Learn module.

Longer, hands-on courses for people who will write or direct code.

Building with the Claude API (Anthropic): Prompting, tool use, RAG, agents, MCP and production patterns. 67 lessons, about 9 hours.
Claude Platform 101 (Anthropic): Building on the Claude Platform from the ground up. About 1.5 hours.
Claude Code 101 (Anthropic): The terminal-based agentic coding tool and its core workflows. About 1.5 hours.

The two topics hiring managers ask about most. See also how to explain them in interviews.

Build with Retrieval-Augmented Generation (OpenAI): Part of the API pathway on OpenAI Academy.
Evaluate AI Applications (OpenAI): Part of the API pathway on OpenAI Academy.
Retrieval Augmented Generation (RAG) (DeepLearning.AI): Listed as intermediate level, a long course of more than 20 hours.
Evaluating and Improving Agentic Applications (DeepLearning.AI): Listed as intermediate level, about 15 hours.

The open standard for connecting tools and data to AI apps.

Introduction to Model Context Protocol (Anthropic): Building MCP servers and clients with the Python SDK. About 1 hour.

Useful if your target employers run on a specific cloud.

Google Skills: agents learning paths (Google): Google describes a no-cost subscription with 35 monthly credits for hands-on labs. Confirm the current terms there.
Microsoft Learn: AI hub (Microsoft): Learning paths for developers, data professionals, IT and business users.
Claude with Amazon Bedrock (Anthropic): API use, tool use, RAG and agents on AWS. About 8 hours.
Claude with Google Cloud's Vertex AI (Anthropic): Setup, prompting, tool use, RAG, agents and deployment on Vertex AI. About 8.5 hours.