[01] Roadmap
Built, in development, planned.
Graphentra is pre-launch. This is what exists, what is being built for the MVP, and what comes after. There are no dates, on purpose.
What the labels mean
- Built · prototype
- Working in the prototype. Prototype grade, not a released product.
- In development
- Being built for the MVP. Not yet available.
- Planned
- Intended for after the MVP. Subject to change.
[02] Now, Next, Later
Every claim, with its status.
This list is the same one that drives the status chips on the rest of the site, so a page cannot say more than this one does.
Now
Built · prototypeBuilt in the prototype
- From changed functions to their callers Built · prototype
Deterministic TypeScript analysis of a Git diff, from changed functions to their callers, using the TypeScript compiler and type checker.
Named functions today. Wider coverage is in development.
- Content-hashed evidence file Built · prototype
A schema-validated evidence file identified by a content fingerprint. Same repository content, same diff and same engine version give byte-identical evidence.
- Works on day one Built · prototype
Starts from the repository and the diff. No test history, instrumentation or recorded sessions are needed.
- Application map Built · prototype
A heuristic map of routes, pages, APIs and end-to-end flows for Next.js, React Router, Express and NestJS.
Heuristic and best effort. Not yet part of the hashed evidence.
- QA checklist written from the evidence Built · prototype
An LLM writes a plain-English QA checklist from the evidence. It narrates the evidence and never changes it.
Needs an LLM today. A report that needs none is in development.
- Local graph explorer Built · prototype
A local, read-only explorer for the dependency graph behind an analysis.
Next
In developmentIn development for the MVP
- GitHub Action with a sticky PR comment and check In development
Runs the analysis in your own CI and posts one comment and one check on each pull request.
- Surfaces inside the hashed evidence In development
Screens, pages and API endpoints become part of the evidence, so surfaces are as reproducible as the code paths that reach them.
- A proof path for every affected surface In development
Each affected surface carries the path that reaches it: changed function, caller, component, route.
Caller paths for changed functions exist in the prototype.
- Proven, likely and unknown confidence tiers In development
Every affected surface is marked proven, likely or unknown. What cannot be traced is listed as unknown, never silently dropped.
- Mapping to existing tests In development
Existing tests that reach the change are listed, and affected surfaces with no test are flagged.
- A complete report that needs no LLM In development
The full report is produced deterministically. LLM narration stays optional.
- MCP server for coding agents In development
Tools an agent can call: impact_of, affected_surfaces, tests_to_run and explain_path.
- Useful, Noise and Missed feedback In development
One click on every report to say whether it was useful, noise, or missed something.
- Wider TypeScript coverage In development
Arrow functions, components, class methods and JSX renders, not only named functions.
- Public accuracy benchmark In development
A public benchmark, ImpactBench, on which recall and precision will be published.
- Hosted report page In development
A web page for each report, with surfaces, proof paths, tests and unknowns.
Later
PlannedPlanned after the MVP
- Release-level regression scope Planned
The pull requests in a release, combined and deduplicated by surface: what to test for the release.
- Run only the affected tests, with a CI-minutes report Planned
Run just the tests that reach the change, and report the CI minutes that saves.
- Incident back-trace Planned
Start from a broken screen and find the recent pull requests that reached it.
- Zero-impact fast lane Planned
Pull requests with no reported surface impact get a clear check, so they can move faster.
- Coverage import Planned
Import test coverage so a static path can be confirmed by what actually ran.
- Compliance and audit export Planned
Reproducible impact records you can attach to a change.
- Generated tests for uncovered surfaces Planned
Draft tests, for your team to review, only for flagged surfaces that have none.
- Isolated browser and API verification Planned
Approved browser and API checks run in disposable environments. Deferred until after the MVP.
- Python Planned
A second language, after TypeScript depth is proven.
- Open-source core Planned
An open-source core of the engine. The licence is undecided.
- Report history Planned
Past reports for each repository, so you can see how a surface has been reached over time.
- Single sign-on Planned
Single sign-on for larger organisations.
The limit of proof. Proof applies to the paths Graphentra reports. No path found is not proof of safety.
[03] Not yet
What Graphentra does not do yet.
- Other languages Python is planned after the MVP. Not yet
- Vue and Angular Not on the MVP path. Not yet
- GitLab and Bitbucket GitHub only for now. Not yet
- Cross-repository analysis One repository, or a simple workspace, at a time. Not yet
- Running tests Graphentra lists the tests that reach a change. It does not run them yet. Not yet