Developer ToolingIn Development
Proofmark
An AI QA agent, in development. It is designed to run your repository in a sandbox, measure existing coverage, and write and run new tests within a set budget. It returns a report, a pull request, and a record of what is still untested.
The problem
AI-generated tests are only useful if every reported test has actually been executed. The agent has to verify its own work and account for what it could not cover.
How it works
A phased run inside a Docker sandbox with a per-phase egress proxy: reconnaissance, baseline, planning, writing, verification. Tests are reported only after execution. Output is a structured report, a reviewable pull request, and a coverage ledger. Each architecture decision is documented in the repository.
Status
The run environment, CLI, report schema, and the reconnaissance and baseline phases are complete and running against two open-source repositories we forked for testing. The planning and test-writing phases are in progress.
Have a similar problem?
Every engagement starts with a fixed-price two-week pilot that runs in your repository and your cloud.
Email us about a pilot