roadmap.sh: AI Agents
A free, interactive visual roadmap for designing, building and shipping AI agents: model basics, prompting, tools, memory, agent loops, evaluation and safety. It helps you understand what agent features in tools and frameworks are actually doing, and which parts you still need to learn.
Roadmap
AI & Machine Learning
roadmap.sh
Self paced
Intermediate
Free
What it is
The AI Agents roadmap on roadmap.sh is a visual learning path for building software where a language model decides which tools to call to reach a goal. It covers the foundations of large language models and prompting, then tool use and function calling, memory, the loop an agent runs, common architectures, evaluation and testing, and the security and safety concerns specific to agents. Each node opens a short explanation with links to further reading.
Who it is for
Developers and automation builders who have started using agent features, such as the AI Agent node in n8n, AI Agents in Make or Zapier Agents, and want to understand what happens underneath. It is also a good map for anyone deciding whether an agent is needed at all, or whether a simple chain of prompts would do the job.
How to use it
Use it as a checklist rather than a course:
- Read the fundamentals nodes on models, tokens and prompting first.
- Focus next on tools, memory and the agent loop, because those explain most agent behaviour and most agent bugs.
- Spend real time on the evaluation and security nodes, especially prompt injection and giving agents only the permissions they need.
Pair it with hands-on practice in your own tool and with provider documentation.
Is it worth it?
As an overview, yes. Agents are a fast-moving area full of marketing language, and a structured map helps you separate core ideas from product names. It will not teach you a particular framework in depth, and explanations are short by design. The roadmap is written by the roadmap.sh community and the project is open source.
Time needed
Skimming takes an hour. Working through the parts that matter for building reliable workflows with agent steps is a few weeks alongside practice.
Common questions
Do I need to code? Not to read it. Some linked resources assume Python or JavaScript.
How does it differ from the AI Engineer roadmap? The AI Engineer roadmap covers the broader job of building products on models, including retrieval and deployment. This one zooms in on agents.
Is it free? Yes, the roadmaps are free to use.
Ready to start?
Opens on roadmap.sh in a new tab.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.