Aviator Verify — Replace Code Reviews with Verified Intent

Aviator Verify is a product that replaces traditional code review with intent-based verification. The core pitch: instead of asking whether code looks okay, ask whether it matches the team's stated intent and expected behavior. Developers capture what they and their AI coding agent agreed to build before opening a PR, and Aviator checks every acceptance criterion against running code with attached evidence.

Verification runs through three layers: end-to-end scenarios that exercise the change and capture screenshots, tool calls, and request/response data; invariants that encode past team review comments as reusable checks so flagged patterns don't recur; and code-scan checks for structural concerns like endpoint existence, return-type matching, and dependency changes. An LLM fallback handles complex criteria with a confidence threshold, and the verdict is labeled accordingly.

The workflow integrates with existing AI coding tools (Claude, Cursor, Copilot) via an MCP server. The agent captures intent, generates acceptance criteria, and submits everything in one tool call. Reviewers then see each criterion with a verdict, evidence, and the ability to run ad-hoc scenarios, waive with reason, or approve. Every verification produces an immutable audit record with full traceability from intent to deployment, targeting SOC 2 and ISO 27001 compliance needs.

The page contrasts Verify against AI code reviewers that "read the diff and post comments" without knowing intent, emphasizing reproducibility, encoded team patterns, and the audit trail as differentiators.