For developers
Get help with your AI coding setup.
You use Cursor, Claude Code, Codex or Copilot. Maybe one assistant, maybe several agents. We help you make that setup safer, cheaper and easier to trust.
Start free: run our check on your own setup. If it finds something, we can review it, or build it with you.
Three ways we can help
1Free setup checkPaste a prompt into your own AI tool. Get a verdict with the file and line behind it.
- Who it is for. Anyone using an AI coding tool, from one assistant up to several agents.
- What you get. The Check for yourself prompt (it sits on every page of the site, and on each lesson), plus the lesson prompts on the free checker page. Your own agent reads your files and answers. It changes nothing. You get one line per idea: already doing it, would help, worth a closer look, or does not apply.
- How long. About 10 minutes.
- Price. Free.
2Setup review, USD 750We read your AI dev setup and send back a written plan.
- Who it is for. A developer or small team that ran the check, or already knows the setup is loose, and wants a second pair of eyes.
- What you get. A written plan, in plain words, that covers six things:
- Rules files: what your agents are told, and where it is missing or contradicts itself.
- Review gates: tests and checks that block a bad change, instead of a prompt that asks nicely.
- A second-model reviewer: someone other than the author checks the work before it ships.
- Memory: what survives a closed chat, and what is quietly dropped.
- Cost controls: which model does which job, and what each costs you.
- Tests: what is covered, and what an agent can break without anyone noticing.
- How long. 3 to 5 business days after we have access to what we need to read.
- Price. USD 750, fixed. The fee is credited in full against a build booked within 30 days.
3Hands-on build helpWe set it up with you.
- Who it is for. A team that has the plan (ours or yours) and wants someone who has done it to build it alongside them.
- What you get. We work in your repo with you. We set up the rules files, the gates, the second-model review, the memory and the cost controls from the plan, and we leave you able to run it. Scope and price are agreed in writing before any work starts.
- How long. Scoped after the review.
- Price. Scoped and priced in writing after the review.
Before you decide
Does DXdev actually do this?
- We run these ideas every day in our own live businesses.
- One person sets direction. A single AI session, the hub, makes the judgment calls and hands work out. Separate agents each cover one area and stay inside it. Cheaper models do most of the work, and the pricey one decides.
- Rules that matter are code, not prompt text: tests must pass before a commit goes through, and some reads are blocked by hooks.
- Memory is a folder of short notes with a one-page index, and a checker that catches an index over the load limit.
- We also say what is not true: the boundaries are written rules plus hooks, not a sandbox.
Questions
- Do I need to buy anything to start?
- No. The check is free and runs in your own AI tool. Only the review and the build cost money.
- Will you see my code?
- Only what you choose to send. The free check runs on your side and sends us nothing. For a review, you tell us what to read, and our data handling page says how we treat what you send.
- I only use Copilot. Is this for me?
- Yes. A shared rules file, tests and a human review before merge help with one assistant too. We match the advice to your level.
- What if my setup is already good?
- Then we say so and charge nothing for saying it. The free check usually tells you which one it is.
- Is this the same as the AI Workflow Readiness Assessment?
- No. That one is for business owners and workflows. This page is for developers and their AI coding tools. If you run a business and not a repo, start at our services page.