Who maintains the app your agency built?
Your quote tool, customer portal or website works until it does not. An AI engineer and a separate reviewer look after it on a server that is yours alone, and nothing changes without your approval.
Custom Permits, in numbers from their case study
180+
permit types handled, across all US states and Canadian provinces
6
Teammates running for them around the clock
700%
more Google impressions in three months, from their SEO Teammate
On their customer portal, the agent proposes and tests changes and a QA agent grades every code change before it goes to your branch rules.
The app exists. Nobody has it as their job.
Most owner-led businesses got their app one of three ways: an agency built it, a sharp employee built it on the side, or someone inherited it. Then the project ended or the person moved on. Fixes wait, small requests pile up, and nobody reviews what changes.
Associates AI gives that app an owner. An Engineer Teammate works in your code repository, and a separate QA Teammate grades everything it changes, and anything that falls short of the grade you set goes back to the engineer. You stay in charge: you describe what you want in plain English.
It is not a replacement for a large engineering team. It is a responsible set of hands for an app that today has none.
An engineer and a reviewer, on a server that is yours.
An engineer for the software you already have
The Engineer Teammate works in your repository on a server that is yours alone. It fixes bugs, makes the changes you ask for and opens a pull request with its test plan.
A separate reviewer
A QA Teammate that did not write the change reviews it. The engineer has to answer every finding.
A QA agent grades every change
Changes that fall short of the grade you set are sent back to the engineer. You do not need to read code to see what changed and what is still uncertain.
One flat price
$100 a month for any number of people. It includes one Standard Small server and 20 GB of storage. Model usage is billed at provider list rates.
How a change gets reviewed.
-
1
You ask
Describe the change in plain English: a bug, a new field, a wording fix. The Product Teammate turns it into numbered acceptance criteria you can read and correct.
-
2
The engineer builds it
It writes a test that fails for the stated reason, then makes it pass, and opens a pull request with the command output.
-
3
Greenlight grades it
A separate QA reviewer grades the pull request 1 to 6 on a fixed rubric. Greenlight approves only at the grade you set, and findings come back to the engineer.
-
4
Short of your grade, it goes back
A change that falls short goes back to the engineer. Your branch rules still decide the merge, and you get a plain-English summary and the list of open questions.
See Greenlight in detail → · See four real trails from our own repository →
A customer reports that the quote form accepts an empty phone number.
An illustration of the process, not a Custom Permits story.
This is the kind of small fix that sits for months when nobody owns the app.
- 1.You send the request in plain English. The Product Teammate writes numbered acceptance criteria, for example: the form refuses an empty phone number and shows a clear message.
- 2.The Engineer Teammate writes a test that fails because an empty number is accepted, makes it pass, and opens a pull request with the command output.
- 3.The QA Teammate reviews the pull request and Greenlight grades it. A finding, such as a second form with the same gap, goes back to the engineer to answer.
- 4.You get a plain summary and the list of open questions. You approve, and the change ships.
You asked in one message and approved in one step. The test and the review stay on the pull request for whoever works on the app next.
Pro, priced at list.
Pro is one flat price for any number of people and includes one Standard Small server and 20 GB of storage. The table shows the engineering-lane example from the pricing page: Performance Medium server, included credit applied.
| Pro plan, any number of people | $100/mo |
| Server: Performance Medium, 2 vCPU / 8 GB | $124/mo |
| Included server credit: one Standard Small server | −$71/mo |
| Disk beyond the included 20 GB: 0 GB | $0/mo |
| Model usage | Billed at provider list rates, depends on volume |
| Fixed monthly before model usage | $153/mo |
Cost ceiling
Model usage draws from a prepaid account balance. With automatic top-up off, your teammates stop when the balance runs out instead of running up a bill. With it on, you choose the threshold and the top-up amount.
Find out what it would take to look after your app.
Thirty minutes. We look at what you have, tell you plainly whether an AI engineer fits it, and show you how a change gets reviewed.
Book a 30-minute look