AI agents work in the demo.Learn to run them in production.
Hands-on live labs for engineers and the leaders who run their teams. Each lab builds one part of a production agent, live, and you leave with the code.
Agents pass the demo. Then they stall.
The four questions that decide whether an agent reaches production. Each one is a week of the buildcamp, and most are a free lab first.
A task runs for a day and the process restarts. Does the agent send the refund twice?
Answered in week 2, buildYour agent demo works. Who is it acting as, and what may it touch?
Answered in week 3, secure and deployThe bill doubled this month. Which agent, which prompt and which customer did that?
Answered in week 3, secure and deployA model update shipped last night. How do you know the agent did not get worse?
Answered in week 4, govern and extend
Built for the people who ship it
Engineers build the system. Leaders decide where it runs and who signs off. The labs are written for both.
- Engineers
Take an agent from demo to production
Tool calling, memory, multi-agent hand-offs, durable runs, identity, gateways and evals, each built live in a lab you can repeat at work.
See the Live Labs - Engineering leaders
Get a team shipping agents safely
Decide which work goes to agents, who approves what they do and how you measure it, then train the team on your own codebase.
See the page for leaders
Every part a production agent needs
Each lab builds one part of the system. Pick one to see where it sits; the buildcamp builds all of it, a column a week.
- Roles and modelsOne job per agent, a model chosen for that job, and a token budget.
- OrchestrationA queue agents pick work from, with hand-offs a person can follow.
- Context and memoryRetrieval with sources, memory across sessions, prompts laid out for the cache.
- Tool callingTyped tools that act in GitHub, Linear and Google.
- MCP serversEvery tool behind an MCP gateway, scoped to the role that needs it.
- Durable executionRuns that resume after a crash and never repeat a write.
- Human in the loopA person approves anything that cannot be undone.
- Identity and permissionsEach agent signs in as itself and acts on behalf of a user.
- LLM gatewayRouting, caching and a budget on every model call.
- Traces, evals and costEvery run traced, scored and charged to the agent that made it.
- Multi-tenancyEach team or customer kept apart, in data and in the bill.
- Policy as codeRules checked on every action, not written in a document.
- Signed skillsNew roles built from reviewed, signed skills.
Free Live LabBuild a Mission Control for AI Agents
Free Live LabBuild a Multi-Agent System with MCP
Free Live LabSecure AI Agents with Guardrails
Free Live LabCut AI Agent Costs with Token Economics
Free Live LabBuild a Second Brain with Claude Code
Free Live LabBuild AI Agent Evals from Production Traces
Start free, then go deeper
An hour, a day or four weeks. Each step builds on the one before, so nothing you build is thrown away.
Build a Mission Control for AI Agents
Free Live LabA board your AI agents work from, with each agent's role, its tools and the cost of every run.
1 hour
Freelive on Maven
Save a free seatDesign Production-Grade Agentic AI Systems
One-day workshopAn architecture for one real workflow at your company that you can take to a design review: roles, models, context, tools, approvals, failure plan and budget.
4 hours
$497paid once on Maven
See the workshopAgentic AI Buildcamp for Engineers & PMs
Four-week buildcampDesign, build and deploy a production agentic AI system in four weeks, ending in a demo day.
16 Nov to 11 DecMonday and Wednesday
$997paid once on Maven
See the buildcamp
Coming up
Live sessions, shown in your timezone. Can't make one? Sign up anyway and the recording comes to you.
Watch a past session
Every Lightning session is recorded. Watch the build, then bring your questions to the next one live.
- Recording60 min
Build an Autonomous Claude Code Agent
Turn a GitHub issue into a task it runs in its own worktree, without you.
Watch the recordingMaven - Recording60 min
Build a Claude Code Verification Harness
Reject code automatically before it merges or deploys, when it breaks your engineering rules.
Watch the recordingMaven - Recording60 min
Build Your First Software Factory Execution Harness
Give each agent task its own Git worktree that is created, used and removed cleanly.
Watch the recordingMaven
“You learn a system by building it, breaking it and fixing it. So that is what every workshop does.”
- Fixes how engineering teams ship with AI agents, at KaaS.team
- 14 years building production software
- Architect of the pivot behind an €11.6M Series A
Questions
What engineers and their managers ask before they pay.
Is there a free agentic AI course?
Yes. Every Live Lab is a free Lightning lesson on Maven, live and recorded, and every past recording is listed here for anyone to watch.
How much does the buildcamp cost?
Agentic AI Buildcamp for Engineers & PMs is $997. A workshop seat is $497. Both are paid once on Maven.
What do I leave with?
The code of the system you built, yours to keep, and a demo day where you show it to people who ask hard questions.
Is it for beginners?
It is for working software engineers who write Python. If agents are new to you, start with a past Lightning recording, then join the next one live.
How long does it take?
A Live Lab is one hour and a workshop four hours. The buildcamp runs four weeks, with two live sessions a week.
What if I miss a live session?
Every session is recorded on Maven for everyone who signed up.
Can my company pay?
Yes. Seats are bought on Maven, which gives you a receipt to expense. For a team, private labs are on the page for engineering leaders.
I lead a team. Where do I start?
With the one-day leaders workshop, or the 2-day audit if you want your own codebase reviewed. Both are on the page for engineering leaders.
Who owns the code I write?
You do. Everything you build in a lab, a workshop or the buildcamp is yours to keep and use at work.