Regression prediction
Trained on your own defect history, the model scores each change by its probability of introducing a regression. Files that have broken before get higher scrutiny automatically.
Four steps from commit to confidence
Most quality gates slow you down. Ours runs alongside your existing workflow and reports back before the pull request is merged.
1
Connect your repository
Point the engine at your Git host. GitHub, GitLab, Bitbucket, Azure DevOps: all supported. The setup takes under five minutes because we use your existing webhooks rather than asking for broad access tokens.
2
Baseline your codebase
On the first run we build a quality profile of your project. This profile records code complexity, test coverage gaps, dependency risk, and historical defect patterns from your issue tracker. Think of it as an X-ray of your repo.
3
Analyse every change
Each new commit triggers a focused scan. The engine compares the diff against the baseline, runs anomaly detection on test outputs, and cross-references known regression signatures. Typical scan time for a 50-file diff: under four minutes.
4
Report and block (or warn)
Results appear as inline comments on the pull request. You choose whether the engine blocks the merge or simply warns. Severity thresholds are configurable per team, per repo, per branch.
What the engine actually does
Six distinct analysis modules run in parallel. You can enable or disable each one independently.
Test gap analysis
The engine maps your test suite to the code it exercises. When a commit changes logic that no test covers, it flags the gap and suggests which test file should be extended.
Dependency risk scoring
Every third-party package in your lockfile is scored for maintenance health, known CVEs, and breaking-change frequency. Upgrades that increase risk surface get flagged before merge.
Flaky test detection
We track test pass/fail patterns across runs. Tests that flip without code changes are isolated and reported with a confidence score so you can fix or quarantine them.
Performance drift alerts
If your build includes benchmark or load-test data, the engine watches for gradual performance degradation across releases. A 3% slowdown per sprint compounds fast: we catch it early.
Code review triage
Pull requests are ranked by risk. Reviewers see a priority queue rather than a flat list, which means the dangerous changes get eyes first instead of sitting in a queue behind trivial edits.
How a fintech team cut escaped defects by 62%
A 45-person engineering org in Edinburgh was shipping bi-weekly. Their post-release defect rate averaged 11 per sprint. After integrating the quality engine into their GitLab pipelines, that number dropped to 4.2 within three sprints.
We stopped treating quality as a phase. The engine made it a property of every commit.
Lead QA engineer, payments platform team
Sprint velocity didn't drop. Merge-to-deploy time actually shortened by 18 minutes on average because fewer builds needed rollback.
62% reduction in 9 weeks
Common questions
It reads diffs and metadata, not your full source tree. The webhook sends us the changed file list, complexity metrics, and test results. If you prefer on-premise analysis, we offer a self-hosted runner that keeps everything inside your network.
The regression prediction and test gap modules work with any language that produces structured test output (JUnit XML, TAP, or JSON). Dependency risk scoring currently covers npm, PyPI, Maven, NuGet, and Go modules. We add new ecosystems based on customer demand, typically within a two-week cycle.
For a repo with 200,000 lines and 18 months of Git history, the baseline completes in roughly 25 minutes. Larger monorepos can take up to an hour. You only run it once; after that, incremental analysis handles each commit in single-digit minutes.
Yes. Each module exposes a threshold config file (YAML) that lives in your repo. Teams can set different rules for different branches. A common pattern: strict blocking on main, warnings-only on feature branches.
Pricing is per active repository per month. We offer a free tier for open-source projects and a 30-day trial for private repos. After the trial, plans start at £120/month for up to five repositories. Volume discounts apply above 20 repos. Contact us for a quote tailored to your org size.
Paid plans include a 99.5% uptime SLA for the cloud-hosted scanner. The self-hosted runner depends on your own infrastructure, but we provide monitoring hooks and alerting templates for Prometheus and Datadog.
Start a conversation
Tell us about your stack and your biggest quality pain point. We will reply within one working day with a concrete suggestion, whether or not you end up using the engine.
2 St Andrews Close, Nether Brekke, Scotland, QU55 7CA, United Kingdom