1
1 Comment

I built a VS Code approval boundary for AI repo work

I built this because I wanted help from AI coding tools without handing them silent control over the repo.

The shape is small on purpose:

  • see the plan first
  • approve once
  • let it run locally
  • keep proof afterward

That is what ApproveOnce does.

It is a VS Code-first product for people who like AI coding speed but still want a real approval boundary before the repo gets touched.

Product:
https://arkheos.ai/approveonce/

Proof:
https://arkheos.ai/proof/

The question I am testing is pretty simple:

will builders pay for tighter control over AI execution, or is this something people only say they want?

If you have built, sold, or used AI dev tools, I would love your read on:

  • whether this feels like a real product or just a workflow
  • whether the trust/control wedge is sharp enough
  • where you would expect the first real buyers to come from
  • whether the current one-time founding-access offer feels early, fair, or wrong
on March 22, 2026
  1. 1

    approval boundaries for AI in repos is exactly the right layer to build right now. AI coding assistants are getting powerful fast but the trust model is still "let it do whatever and hope for the best." having a hard gate where certain files or directories require human approval before AI can modify them is smart. we use a similar pattern in our automation — hooks that block certain actions and force a review before they execute. works well for preventing AI from accidentally modifying config files, secrets, or production code. is this VS Code specific or could it work with other AI coding tools?