> 01
Foundations
Prompting, tool calling and structured output: the building blocks every agent is made of, and how they fail.
> For software developers building with AI
Practical tutorials, in-depth guides and starter kits that take you from your first tool call to agents you can trust with real users. Start free. Go deeper when you're ready.
1// Step 4 — give the agent tools, then a budget.2// Unbounded loops are how demos become bills.34const agent = createAgent({5 tools: [searchDocs, openTicket],6 maxSteps: 8,7});8await agent.run(task);What you'll learn
Whether you've never shipped an LLM feature or you're already running agents in production, there's a next step for you. Every lesson is hands-on: you write the code, run it, and break it.
> 01
Prompting, tool calling and structured output: the building blocks every agent is made of, and how they fail.
> 02
Agent loops, planning, multi-agent setups, and knowing when a simple workflow beats an agent.
> 03
Designing tools a model can use well, connecting APIs and MCP servers, and keeping permissions tight.
> 04
Retrieval, context engineering and memory that holds up as conversations and codebases grow.
> 05
Testing non-deterministic systems: evals, guardrails, tracing, and catching regressions before users do.
> 06
Latency, cost, security and deployment: everything between a working prototype and a product.
The library
The fundamentals are free for everyone. When you want to move faster, our guides and starter kits give you the depth and the code to do it.
Tutorials
FreeStep-by-step builds you follow at the keyboard. Each one starts from an empty file and ends with an agent that runs.
Guides
PaidLong-form playbooks on the hard parts, like evals, memory and production hardening, with the trade-offs explained.
Starter kits
PaidProduction-ready templates and repos you can fork and ship: agents, tools and eval harnesses, wired up and documented.