A pilot for your team
Use your own development work to assess whether Anvil Coder fits your team. Start with one repository and five small tasks. Agree what success means before the evaluation, then inspect the changes, executed checks, review effort and costs.
Select small maintenance changes, regression tests or narrowly defined fixes in one repository. Each task needs acceptance criteria and a baseline: how would your team handle it with its current tools?
Confirm that the language, dependencies and build environment can be supported. Use test data approved for the evaluation. A task with unresolved product decisions or unavailable checks should be clarified before it enters the pilot.
For each task, record the diff, checks actually executed, failed attempts, human review time and incurred costs. A generated test is an artifact; an executed test with a recorded result is evidence.
Assess accepted changes and the work needed to reach them together. Keep unsuccessful attempts in the comparison. The result should help you decide whether to continue, address missing prerequisites or stop.
Before starting, establish who can grant repository access, which model provider may process the source code, who reviews the changes and how costs and support will be handled. Agree scope and dates for your evaluation.
The platform is hosted in Switzerland. Source code included in model requests is processed by the configured model provider; the complete data route needs to be reviewed for the selected application. The pilot proposal is a starting point for that discussion.