1. Mapping and learning your system.
Connect Assay to your repo. It walks the codebase and maps your entire system: services, modules, data stores, queues, environments, and how they connect.
It also learns how to operate each part: how to boot it, seed it, call it, and where its effects show up. Your team corrects and accepts the map until it’s accurate.
From then on, engineers and agents share one language and a clear mental model of how the software actually works, so every agent starts from the right entry point with the right data.
2. Context gathering for your team.
Tasks can start as Linear tickets, Slack threads, GitHub Issues, or just PRs.
Assay bundles the task context from those places with the code change itself: the business goal, the spec, evolving requirements, options considered, implementation decisions, and how the change should be verified.
Assay stores that context in your repo, so the next engineer or agent can understand the work without digging through old chats or trying to reconstruct what happened on the spot.
3. Running the change in a sandbox.
Assay helps your team choose the behaviors and scenarios to check.
Then Assay spins up the real system using the map: starts the services, seeds the data, and drives each scenario through a real entry point like an API call, a UI action, a queued job.
Every step is recorded as it happens: the request, the database write, the message hitting the queue, the log line, the final state. Open any step and inspect exactly what happened.
4. Shareable visualization and proof for your team.
Assay turns the sandbox run into a visual review on your platform. A reviewer opens one page and sees why the change exists, the code it touches, the scenarios checked, and the run timeline with evidence at every step.
They can rerun a scenario with their own data, challenge a result, comment, request changes, or approve.
Each review is saved as a shareable artifact and attached back to the PR, so the team can see what worked, what failed, what was reviewed, and what was approved or rejected, and why.