Train your team to build with Claude and Codex.
A hands-on program that teaches your people to develop real products and automate real workflows using agentic AI — the same method we use to deliver client work. Your team leaves having shipped something, not having watched a demo.
Delivered on-site or remote, in English, across the US, UK, EU, UAE and ANZ. Curriculum is tailored to your stack and codebase before day one.
What your team can do afterwards
Engineers who direct agentic AI instead of racing it
A repeatable spec → build → review workflow your team owns
Review, testing and security standards for AI-written code
A real internal tool, built during the program and running afterwards
Prompting is not the skill. Delivery is.
Teams that plateau with AI usually treat it as a faster autocomplete. The teams that pull ahead treat it as capacity that has to be specified, directed and reviewed — which is a management problem as much as a coding one.
Specify, don't chat
A written spec with acceptance criteria is what makes agent output reviewable. We teach people to write one, and to hold the agent to it — the same artifact then documents the system.
Run work in parallel
The compression comes from several agent workstreams running at once against the same spec, not from one person typing faster. That takes a workflow, and the workflow is most of the training.
Review like it matters
AI-written code that nobody understands is a liability. Teams leave with a review standard, a test bar and a security checklist they'll actually apply on Monday.
Three tracks, three audiences.
Run one, or run all three so the engineers, the process owners and the leadership come out of it speaking the same language.
The core craft: driving Claude Code and Codex across a real codebase — planning, specs, tests, refactors and the review discipline that makes agent output safe to merge.
Modules
- How agentic coding tools actually work — context, tools, planning loops
- Writing specifications an agent can be held to
- Driving Claude Code: projects, skills, subagents and parallel workstreams
- Driving Codex alongside it, and choosing the right tool per task
- Connecting your systems with MCP so agents work on your real data
- Reviewing, testing and hardening AI-written code
For the people who own the process rather than the codebase: turning manual, repetitive work into automated workflows and internal tools — safely, and without waiting on an engineering queue.
Modules
- Spotting the workflows worth automating — and the ones that aren't
- Describing a process precisely enough for an AI agent to build it
- Building an internal tool end to end with agentic AI
- Connecting spreadsheets, email, CRM and internal systems
- Human-in-the-loop design: where a person must stay in the decision
- Knowing when to hand it to engineering instead
What changes when a team builds this way: throughput, team shape, review load, risk and cost — plus the policy and governance you need in place before you scale it across the organisation.
Modules
- What agentic delivery does to estimates, team shape and hiring
- Governance: policy, approved tools, data boundaries and audit
- Security and IP posture when AI writes production code
- Measuring it honestly — throughput, defect rate, review load
- A staged rollout plan for your organisation
Run it the way your team can absorb it.
Every format is hands-on and built around your own codebase and backlog. Nobody sits through a slide deck about what AI might one day do.
Workshop
One to three days, on-site or remote, for a single team. The fastest way to find out whether this changes anything for you.
Cohort program
Four to six weeks part-time across several teams, each shipping a real internal tool with our engineers reviewing the work.
Embedded enablement
Our engineers work inside your team on live delivery and train as they go — the fastest transfer, and the only one that survives contact with your actual codebase.
We teach the method we ship on.
Nirvana builds and operates five live platforms of its own, all delivered this way — spec first, agentic AI directed by senior engineers, nothing merged unreviewed. The training is that working method written down, not a course someone assembled from blog posts.
- Taught by the engineers who deliver client builds, not by trainers
- Curriculum tailored to your repositories, stack and tooling
- Your team ships a real internal tool during the program
- Governance, security and review standards included, not sold separately
AI Builders, answered.
Who is this training for?+
Three audiences, and they learn different things. Engineers learn to drive Claude Code and Codex across a real codebase — specs, tests, refactors and the review discipline that makes agent output safe to merge. Operators, analysts and process owners learn to automate their own workflows and build internal tools without waiting on an engineering queue. Leadership gets a one-day workshop on what changes in throughput, team shape, risk and governance when a team builds this way.
Do we need to be developers to take part?+
Not for the workflow track. It's built for the people who own a process rather than a codebase — they learn to describe a process precisely enough for an AI agent to build it, and where a human has to stay in the decision. The builders track does assume working software engineers.
Is this vendor training for Claude or for Codex?+
Neither, and both. We teach the method — specification, parallel agent workstreams, review, testing and guardrails — using both Claude Code and Codex, including when to reach for which. We are not resellers of either; the tools change, the method transfers.
What does the team actually build during the program?+
Something real from your own backlog. We pick a genuine internal tool or workflow before the program starts, and your team ships it during the sessions with our engineers reviewing the work. That is the point — a team that has shipped once with the method will do it again, and a team that has only watched a demo will not.
How do you handle security, IP and our source code?+
We work within your policy, and setting that policy is part of the leadership track. We cover approved tools and commercial API terms, data boundaries, keeping secrets and production data out of agent context, dependency scanning and licence checks, and the review and audit trail that has to sit around AI-written code before it reaches production. If you have no policy yet, you will leave with a draft.
Can you train on our own codebase and stack?+
Yes — and it's the version that works best. We tailor the curriculum to your stack, repositories and tooling before day one, and the embedded format puts our engineers inside your team on live delivery so the transfer survives contact with your actual code.
How is this different from the AI training we already tried?+
Most AI training teaches prompting. This teaches delivery: how to specify work an agent can be held to, how to run several agent workstreams in parallel, how to review and test what comes back, and where a human must stay in the loop. It's the same method we use on our own client builds, taught by the engineers who use it.
How do we get started?+
Tell us the team, the stack and the workflow you would most like to see automated. We come back with a tailored curriculum, a format recommendation, and the internal tool your team would build during the program.
Let's build
What's your AI Nirvana?
Tell us where you want to go. We'll bring the team, build the product, and grow it with you — and you own it.
- You own the IP
- US-based team
- Reply within 1 business day