Build a working Claude agent
in six weeks.
A hands-on cohort for intermediate engineers. By week 6, you've deployed a real agent solving a real problem at your business — code reviewed by a Nolola partner, certified, and (if you're hireable) eligible for the AI Talent Marketplace bench.
Capstone-required. You don't leave with a PDF certificate that means nothing. You leave with a deployed agent your team is using.
Engineers who want
to actually finish.
Cohort structure exists because most self-paced courses get abandoned in week 2. This one ends in week 6, with a working agent.
- You can read TypeScript or Python and have shipped to production before
- You've called the Anthropic API but haven't built a real agent yet
- You've watched YouTube tutorials and want something that actually finishes
- You have an agent idea for your business and need 6 weeks of structure to ship it
- You want a Nolola-verified credential, not a self-attested LinkedIn skill
Never called the Anthropic API? Start with the free Prompt Engineering Basics first. Want the agent built FOR you instead of by you? See the Expert Marketplace.
A real agent.
Not a toy.
Everything in your repo, on your infra, solving a problem you pick in week 1.
A real Claude agent
Not a toy. The agent solves a real problem at your business — internal triage, customer Q&A, ops workflow, sales enablement, whatever you scope in week 1.
Production-grade tooling
Observability + evals wired up. Cost tracking. Versioned prompts. Cold-start tested. The full deployment kit, not the demo kit.
Code you own
Everything is in your repo on your infra. No Nolola lock-in. The capstone is yours to extend, modify, or scrap.
A Nolola-verified certificate
Capstone reviewed by a partner before issuance. The certificate means you shipped a working agent — not that you watched 6 hours of video.
Week by week.
Ends in production.
The agent loop
What an agent actually is — a model with a goal, tools, and the ability to decide when to use them. Build the simplest possible agent by lesson 2.
Tool use + structured output
Claude's tool-use API, structured output, JSON-schema constraints. Build the toolbox your capstone agent will reach for.
Memory + state
Short-term memory, retrieval-augmented memory, persistence between turns. The difference between a chatbot and an agent.
Multi-step reasoning + planning
Plan-then-act, reflect-and-revise, when to break a job into sub-agents vs one looped agent. Real production tradeoffs.
Production readiness
Observability (Langfuse, Arize), evals, cost control, failure modes, prompt versioning. The unglamorous half of shipping.
Capstone deploy + review
Final agent goes live. Code review with a Nolola partner. On approval, you get certified and (if you're hireable) an invite to the AI Talent Marketplace bench.
Learn it,
or hire someone who has.
Building AI Agents with Claude
- Deliverable
- Shipped capstone agent + Nolola-verified certification
- Right fit
- You want a real working agent + structure to ship it
Shipped in production.
Still running today.
“Came in having called the API. Left with a working triage agent in our support stack, code reviewed by a partner, certified, and a slot on the marketplace bench. The reviews caught things I would have shipped wrong.”
“I'd been stuck at ‘works locally, breaks in production’ for months. Week 5 unlocked it — evals, observability, cost control. The agent has been running for our team since week 6.”
“I'm a non-technical founder, took it with our lead engineer. We co-built the capstone. He learned Claude, I learned what to ask for. Now I can scope agent projects without him having to translate.”
Common questions
Intermediate. You should be able to read TypeScript or Python and have shipped something to production before. If you've never called the Anthropic API, start with the free Prompt Engineering course on /academy first.
You keep all course materials and Slack access. You just don't certify on time. If you ship within 90 days of cohort end we'll review it and certify retroactively — most students who don't finish on schedule do finish eventually. About 75% of starters complete and certify.
30-minute code review with a Nolola partner who's shipped agent systems in production. We grade on: does it work end-to-end, does it have evals, does it handle failure, would you ship this to a real user. We're happy to fail you if it's not ready — that's what makes the certificate mean something.
Plan for ~5-7 hours: 1 live session (90 min), 2-3 hours on assignments, 1 hour in Slack and reviewing peers. Weeks 5 and 6 are heavier as the capstone comes together — closer to 10 hours.
Two reasons. (1) Claude's tool use, structured output, and 200k context window make agent workflows more predictable than the GPT equivalents at the time of writing. (2) Most of what you learn transfers — the agent loop, evals, observability, planning patterns are model-agnostic. You'll come out able to swap models with a config change.
Yes — that's the point. The capstone is supposed to solve a real problem at your business. Bring it in week 1, scope it together with your mentor in week 2, ship it by week 6. If you need help getting it past legal/security, we've done that before with several students.
Top-performing certified grads get an invitation to apply for the Nolola AI Talent Marketplace, where companies actively hire AI builders for paid engagements. Not auto-listed — you still go through expert vetting — but the certificate puts you in the funnel.
Yes. 4+ seats from the same company get 15% off and a kickoff call with the cohort lead to align capstone scope with your internal AI roadmap. Email hello@nolola.com after the first seat is booked.