[01] Use cases
Three people, one question.
After a change, someone has to decide what to test. These are the people who make that call, and what Graphentra is designed to give each of them.
Usually the first to feel it
Test what the ticket did not mention.
01 · QA teams
A QA lead is handed a ticket titled “Update token validation to check the audience claim”. It names one flow: OTP login. The change is to a token validator that password login and guardian access also use. Nothing in the ticket says so.
What Graphentra is designed to give you
- A ranked, plain-English checklist of the screens and flows a change can reach. Built · prototype
- The reason and the proof path beside every item. In development
- Affected surfaces that no test reaches, flagged for a manual check. In development
- A plain list of what could not be traced. In development
Usually the one who signs it off
Review AI-written pull requests on evidence.
02 · Engineering leaders
A pull request titled “Update shared authentication” changes one function. Before merge, the engineering manager wants to know how far it reaches, which of those surfaces a test covers, and what could not be traced.
What Graphentra is designed to give you
- One check on every pull request, produced in your own CI. In development
- Reach, test coverage and unknowns at a glance. In development
- Evidence that is identical every time it is run. Built · prototype
- Less wasted test time, by running only the tests that reach a change. Planned
Where the volume is coming from
Give your agent something to check against.
03 · Developers with AI agents
A coding agent is asked to tighten token validation. Before it edits validateToken(), it asks what the function reaches. After the edit it asks again, and finds a surface that no test covers.
What Graphentra is designed to give you
- impact_of before an edit, and affected_surfaces after it. In development
- A proof path an agent can read back with explain_path. In development
- The tests that reach a change, from tests_to_run. In development
- Evidence designed to feed AI code reviewers as well as agents. In development
[05] Also planned for
Two more, later.
These come after the MVP. They are mentioned so you know where the evidence file is heading, not because they exist.
-
Regulated teams
Reproducible impact records to attach to a change.
Planned -
Agencies
Per-release impact for clients.
Planned
The limit of proof. Proof applies to the paths Graphentra reports. No path found is not proof of safety.