Automated Code Verification With an Audit Trail
Every CodeChecks run produces a cryptographically-linked evidence bundle — repo SHA, patch hash, verification results, and reproduce instructions — so you always know exactly what was checked and why it passed.
The verification pipeline
Three sequential phases, each gated on the previous, produce a result you can trust and reproduce.
Structured task decomposition
Before writing code, the Planner reads your repo structure, identifies affected files, and produces a task list with verification commands. Nothing is written until there is a plan.
Policy-checked unified diff
The Implementer generates a unified diff for every task. Sensitive paths — env files, auth modules, payment code — are blocked by policy and cannot be modified without explicit opt-in.
Independent cross-validation
A second AI model (GPT-4o) reviews the diff independently, returning a structured verdict: pass, fail, or needs-review. The result and its reasoning are stored in the evidence bundle.
Every run ships with an evidence bundle
A CodeChecks evidence bundle is a structured JSON record that captures everything needed to reproduce, audit, and challenge a verification result. Your ops team, security reviewer, or future self can open it and know exactly what happened.
Try it free// evidence.json
{
"repoSha": "a1b2c3d",
"patchHash": "sha256:e4f5...",
"verifierVerdict": "pass",
"verifiedAt": "2026-06-20T...",
"plannerModel": "claude-opus-4-6",
"verifierModel": "gpt-4o",
"reproduceSteps": [...]
}
Repo SHA
Exact commit your code was generated against.
Patch hash
Cryptographic fingerprint of the generated diff.
Verification commands
The commands run to confirm correctness.
Environment snapshot
Node version, OS, and dependency lockfile state.
Step-by-step reproduce
Instructions to replay the run from scratch.
Model metadata
Which AI model produced each stage of the output.
When teams use automated code verification
From scrappy startups to regulated enterprises, the need to trust AI-generated code is universal.
Pre-merge verification gates
Require a CodeChecks verification pass before merging. Teams replace time-consuming manual review of AI-generated diffs with an automated, reproducible gate.
Compliance and audit trails
Regulated industries need proof that code changes were reviewed. Evidence bundles give you timestamped, tamper-evident records for every AI-assisted change.
Confidence in AI-generated code
AI coding assistants produce code fast — but developers are rightly sceptical. Automated verification gives your team a second opinion before the diff lands in main.
Continuous quality assurance
Integrate CodeChecks into your workflow so every change is verified, not just the ones that feel risky. Consistent process beats ad-hoc review.
Run your first automated code verification today
Free plan includes 10 task cycles per month. No credit card needed.