AI you can trust, because you can verify it.
We build tools that make AI prove its work. We help organizations adopt AI with structure, governance and evidence.
Mission
Make AI-assisted work verifiable.
AI can write good code today. It cannot run a delivery. It brings no team structure, no agile discipline, no governance, and none of the judgement that comes from years of running real programs. And it cannot vouch for its own work: "the tests passed" means nothing until a program proves it.
Hevelius brings that structure. Agile Dev Team gives AI agents the roles, the order and the controls of a professional delivery team, and records proof at every step. Our advisory work brings the same discipline to organizations adopting AI, so every result stands up to a board, an auditor or a customer.
Product
Agile Dev Team
Version 0.1.0 · early access
A plugin for Claude Code that works as a complete delivery team. You describe what you need in plain English, and our Agile Dev Team of autonomous agents plans the work, builds it, checks it, and delivers one pull request per change: a packaged set of code changes, ready for you to review and approve.
Every pull request opens with the evidence: each requirement, the test that covers it, and the result. Programs write that record. The agents that did the work cannot change it.
User story E1-S1 Add two numbers Criterion Test Before After E1-S1-1 test_adds_integers fail pass E1-S1-2 test_rejects_text fail pass Quality gate lint, types, tests green Scope check task files only pass Audit round 1 APPROVED
How it works
-
Set up.
Run
/dev:initin your project. It proposes the settings, proves your tests run, and asks before it changes anything. -
Say what you need.
Describe the change in plain English. The analyst agent turns it into user stories: small, testable pieces of work the team can build.
The analyst arrives in version 0.2. In 0.1.0 you write the requirements yourself, starting from a guided example.
-
Get proven work.
Run
/dev:runwith the user story's number. You receive a pull request with the evidence at the top. You review it. You merge it.
| Role | What it does | Status |
|---|---|---|
| Manager | Plans each user story, assigns the tasks, audits the result. Never writes code. | In 0.1.0 |
| Senior developer | Builds tasks that need judgement or span several files. | In 0.1.0 |
| Junior developer | Builds small, tightly specified tasks. | In 0.1.0 |
| Tester | Writes each task's tests before the code exists, then reviews the code. | Planned, 0.2 |
| Analyst | Turns what you say in plain English into requirements the team can build. | Planned, 0.2 |
The rules it works by
- Programs check. Agents do not vouch. No agent is ever the source of "the tests passed".
- Tests fail first, then pass. Code is never accepted on a pass alone.
- Every loop has a limit. When the limit is reached, the team stops and reports.
- No agent merges. Nothing reaches your main branch without you.
Who it is for
- Solo builders and small teams on Claude Code who want every change delivered as a pull request they can review.
- Product owners and founders who know what they need but do not write code.
- Teams that need a clear record of how every change was tested.
What you need
Claude Code, git, a GitHub repository, and Python 3.11 or newer. It runs on Windows, macOS and Linux.
Version 0.1.0 builds one user story per run and works with GitHub. The repository is private during early access.
Advisory
AI adoption, from first assessment to working practice.
For organizations that want AI to deliver, and need to know where to start, what will change, and how to govern it.
-
AI readiness assessment
A clear view of where you stand: use cases, data, skills and controls. You receive a short written assessment with priorities, and a direct answer on what to do first.
-
Transformation
We turn a chosen use case into a working process: the operating model, the roles, and the path from a small pilot to wider use.
-
Strategy and governance
AI strategy, the business case, and governance mapped to published frameworks such as the NIST AI Risk Management Framework. We provide evidence and advice, not certification.
Who you work with
Hevelius AI Group is led by Christopher Zaczek. He has 15+ years in financial services transformation and consulting with seven global banking institutions and a Big Four firm. Working from financial centers including Hong Kong, Singapore and London, he advised C-level executives and delivered projects across EMEA, APAC and the US.
Contact
Get in touch.
Write to us for early access to Agile Dev Team, or to discuss an assessment or advisory work. Tell us about your team and what you want to achieve. We reply by email.