Frequently asked questions

Yes. Senix has a permanent Free plan for one repo and 30 reviews per month. Starter, Team, and Pro are paid monthly plans. See the pricing page for details.

No. We persist the structural-diff metadata — added, modified, and removed symbols with file names and line ranges — and the generated summary. We do not persist your raw source files.

PR metadata (title, author, file counts), the structural diff, and the analysis result. The diff content is sent to the LLM provider for a single request and not retained for training. Account data covers your GitHub identity and installation settings.

Not today. The pipeline is open source if you want to read or fork it, but the hosted product is the supported path. Reach out via the Feedback button if you have a hard self-host requirement.

DeepSeek is the default analysis provider — it has the most reliable structured-output support at low cost. Anthropic, Gemini, and Groq are also supported as alternate providers.

The current prompt scores 94% on our internal eval set. We publish accuracy with each prompt revision and revert on regressions. Calibration drift is real — send us examples that look wrong and we fold them into the next eval run.

Copilot's review suggests line-level edits. Senix answers a different question — what behaviorally changed and how risky is it — in a 3-sentence summary with a fixed risk taxonomy. It is built for teams shipping AI-generated code who need a fast read on blast radius, not a list of nits.

Not yet. The 8-flag taxonomy is fixed today so the analysis stays consistent and comparable across PRs. Custom risk-flag configuration is planned for the Pro tier.

Not yet — it's on the roadmap. See the API referencefor what we're planning.