Agentic AI · Guides
How AI agents actually work, one building block at a time. Every guide has code we ran on a laptop, the real output, and an honest section on where the idea breaks.
A chatbot answers from what the model already knows. An AI agent is a language model running in a loop, where it can also do things: search your documents, call an API, run a test, edit a file. It then reads the result and decides what to do next.
Think of the model as the brain and the tools as its hands. Your code is the body that connects them, and it decides what the hands are allowed to touch.
The guides below take each piece apart: how an agent finds the right information (RAG), how it asks your code to run a function (tool calling), which coding agents you can use today, and what they cost.
Hands-on·11 min read
What retrieval-augmented generation is, why chatbots need it, and a free local RAG pipeline in plain Python: keyword search, embeddings, a hybrid, a 10-question test, and a local model that answers only from what it retrieved.
Read the guide
Hands-on·12 min read
How a language model asks your code to run a function, built step by step in plain Python with a free local model: tool schemas, argument validation, an agent loop, a 15-question accuracy test, and how OpenAI, Claude, Gemini and MCP do the same thing.
Read the guide
Beginner·8 min read
Six coding agents side by side: what each costs, which have a free route for students, which are open source, how they read AGENTS.md, and how far you can trust their permission modes.
Read the guideHands-on·14 min read
What LangGraph is, how it differs from LangChain, and five small Python programs that build a support agent as a graph: state and reducers, the agent loop, memory with a checkpointer, a human approval step with interrupt(), and the recursion limit. Every output on the page is real.
Read the guideProject ideas·15 min read
Thirteen agentic AI project ideas from beginner to advanced. Each names its tools, the part where agents actually fail and a test that proves it works. Includes how crowded each idea is on GitHub and a plain-Python starter agent with an 11-question scorer, run on a free local model.
Read the guideHands-on·16 min read
What a multi-agent system is, the six architecture patterns behind almost all of them, and when several agents beat one. We ran four designs on the same job-offer task with a free local model and measured accuracy, model calls and text sent, including what happens when a subtask loses one detail.
Read the guideHands-on·15 min read
How agents remember: short-term and long-term memory, the four memory types (working, episodic, semantic, procedural), and the five designs behind every memory system. We built all five in plain Python and tested them on 24 sessions of chat where four facts change.
Read the guideHands-on·16 min read
What context engineering means, how it differs from prompt engineering, why long prompts degrade (context rot), and the techniques agents use: clearing, compaction, notes, sub-agents and cache-friendly prompts. We built six strategies in plain Python and measured what each sends the model across one 30-step debugging run.
Read the guideHands-on·17 min read
What prompt injection is, direct and indirect attacks with examples, how it differs from jailbreaking, why agents raise the stakes (the lethal trifecta), and the guardrail types that help. We put four guardrail layers around a small support agent and ran 24 attacks and 24 genuine tickets through them, assuming the model is fooled every time.
Read the guideFree mini courses and tool guides that go deeper on agent building.
Crio AI Engineering
Crio's AI Engineering program is project-first: you build LLM apps, RAG pipelines and tool-using agents, then learn to evaluate, secure and ship them. Talk to the team about whether it fits you.
Last updated