Your system, your data, your deliverable.
All three are named as yours in the contract before any work begins, and they go back to you in a state your own people can operate. Nothing is held back to make leaving hard.
The method page explains how we attach to a system we did not write. This page answers what the engagement is, and what happens the day it ends.
A bounded first look at your system. What it covers is agreed in writing before anything starts, and it does not commit you to anything after it.
The build, if you want one. We build inside the comprehension the first piece produced: in your repositories, under your accounts, through the review your team already uses. Scope and what done means are settled before we write anything.
Ongoing operation, if you want that. Hosting, monitoring, dependency upkeep, new work on a cadence you set. It is something you buy on purpose, never a subscription that switches itself on.
Letting an outside team into your system usually means part of it stops being legible to you, and leaving gets expensive on purpose. So we write the exit down before the work starts, not when you ask for it.
All three are named as yours in the contract before any work begins, and they go back to you in a state your own people can operate. Nothing is held back to make leaving hard.
The tooling and the method we bring in are ours and stay ours. You are not renting our internals, and you are not buying a dependency on us.
What we take away is how we work. What we learned about your business goes to you in writing. It does not go into anybody else's project.
If you want your own team to carry it, they can: what we hand over when we detach is written down, and your own people can operate it without us. If you would rather we kept running it, that is a separate agreement you opt into. Being quiet about this is how a handover turns into a bill nobody agreed to.
Most operations are already half described in a project chat with somebody's own model. Our intake protocol is a plain text file you paste into that chat. It interviews you one question at a time and produces a spec bundle you read before you send anything. It is free and there is nothing to sign.
↳ the protocol, exactly as it runs.
↳ an illustrative bundle.
We have not yet attached to an outside company's system for hire, so there is nothing we could honestly show you there. The method is proven on our own work and offered outward. Where we do have clients, their work is theirs to publish, not ours.
What stands behind it is thirty years of building and running production software, written down as gates a machine enforces. You can read those gates: bonfire-ai on PyPI is the engine we build with, and the open repositories carry them inside.