Deterministic review signals for teams shipping high-stakes changes
Merge confidence for critical releases.
ArcGuard reacts to merge request events, runs a shared analysis engine, and posts a structured Merge Confidence Report directly into GitLab. The result is a calmer review loop with inspectable evidence instead of vague AI opinion.
A calmer interface for reviewing risky change.
The landing system now mirrors ArcGuard's product positioning: minimal chrome, strong hierarchy, and enough atmosphere to feel premium without turning the interface into decoration.
GitLab-native workflow
ArcGuard is triggered from the actual review loop, reacts inside CI, and posts its merge confidence report back into the merge request.
Deterministic analysis engine
Architecture drift, review zones, intent alignment, flaky evidence, rollback safety, and CI waste are scored from structured, inspectable logic.
Demo and flow parity
The landing experience and the GitLab flow use the same shared engine, so the live demo stays anchored to the real product behavior.
Six evidence lenses, one final merge verdict.
ArcGuard is not a generic chatbot layered on top of a dashboard. It is a defined review mechanism with clear inputs, computed outputs, and reviewer-facing evidence that can be challenged or verified.
Architecture drift
Detects risky boundary crossings, elevated dependency exposure, and change shapes that deserve design attention.
Review complexity
Maps the semantic spread of a merge request so reviewers can spot where attention is thin or overloaded.
Intent alignment
Compares what the MR claims against code, tests, and docs so delivery intent stays coherent.
Flake witness
Carries seeded CI evidence into the scoring model rather than treating reliability as an afterthought.
Rollback reality
Checks whether the change is actually recoverable, not merely reversible in theory.
CI waste lens
Scores avoidable runtime, duplicate jobs, and reclaimable emissions as part of merge confidence.
Trigger to action flow, expressed with less friction.
The design language is cinematic, but the product path stays operationally direct: trigger ArcGuard, compute the evidence, and return a verdict teams can act on.
Trigger
Mention or assign ArcGuard when a merge request needs an authoritative confidence read.
Analyze
The shared engine ingests MR context and computes evidence across six review dimensions.
Report
GitLab receives a structured Merge Confidence Report with verdict, score, and reviewer guidance.
Built to demo clearly and ship credibly.
The repo stays honest about what ArcGuard is: a GitLab-native merge confidence system with shared analysis logic, seeded demos, and a clean handoff into evaluator review.
Review merges with the atmosphere of a premium product and the rigor of a scoring engine.
The homepage now sets the tone; the demo route still carries the real analysis logic. Move into the live console to inspect seeded scenarios and the full Merge Confidence Report.