FAQ
The questions worth answering before you get in touch.
Cost, lock-in, your data, messy legacy code, and what agents can actually do — answered straight, including the parts most pitches skip over.
- How much does an engagement cost?
- There is no fixed price list, because the work ranges from a few days standing up a Claude Code setup to a longer build across several teams. Engagements are scoped to a specific outcome and quoted before any commitment, and most start with a small, bounded pilot so you can judge the value before spending more. You will get a clear figure for the work in front of us, not an open-ended retainer.
- What happens when you leave — am I locked in?
- The opposite is the goal. Everything is built in your repository using your tools: prompts, agent roles, skills, MCP servers, and CI hooks, all documented alongside the code. The hand-off includes the reasoning behind each choice so your team can run and extend the setup without me. There is no proprietary platform to keep paying for and nothing that stops working when the engagement ends.
- What data leaves my environment?
- The agent workflows run against your codebase inside your own environment and tooling. Your code is sent to whichever model provider you choose to use (for example Anthropic's API for Claude) under that provider's terms, exactly as it would be if your engineers used the tool directly. I do not retain copies of your code, and any access I need during an engagement is scoped to what the work requires and removed afterwards. If you have specific data-handling or compliance constraints, raise them at scoping and we design around them.
- My codebase is large, old, and messy. Is that a blocker?
- No — that is the normal starting point, not a disqualifier. A demo on a clean toy repo proves little; the work is judged against real CI, real review gates, and the legacy modules nobody wants to touch. Messy code usually means more value from agents on audits, dependency upgrades, and well-scoped refactors. We pick a contained piece of it for the pilot rather than trying to boil the ocean.
- How does a first engagement start?
- It starts with a short scoping conversation about your codebase, your constraints, and what a good outcome looks like. From there we pick one real piece of work — a refactor, a feature, or a recurring chore — and run it through an agent workflow end to end, so the value is visible before you commit to anything larger. If agents are the wrong tool for your situation, you will hear that at the scoping stage.
- What can AI coding agents realistically do — and not do?
- Agents are good at well-scoped, repetitive, verifiable work: scaffolding, mechanical refactors, test coverage, dependency bumps, and first-pass reviews against clear acceptance criteria. They are not good at owning ambiguous product decisions or architecture calls, and they still need human review gates to be safe to merge. The setups I build keep people on the judgement calls and hand the grunt work to agents — not the other way around.
- Will this replace my engineers?
- No. The aim is to raise the throughput of the engineers you have, not to remove them. Agents take on the repetitive work so your team spends more time on design, review, and the decisions that need human judgement. Teams that try to take humans out of the loop entirely usually end up with more review work, not less.
- Do you have case studies or references?
- Named case studies are being added as clients agree to be referenced, so you may not see a wall of logos yet. Until then the proof is in the approach and in the working setup you walk away with: something running in your repo that your team can inspect and extend. If references matter for your decision, ask during scoping and we will sort out what can be shared.
Still have a question that is not covered here?
Tell me what you are working on and what is in the way. I will respond with whether I can help and what a sensible first step looks like — no boilerplate sales call.
Start a conversation