Veris

If you're running AI coding agents and want to know what behaviors a PR puts at risk before merging, this is the verification layer. It parses your TypeScript or JavaScript repo into a behavioral graph, diffs it against your base branch, scores blast radius and fragility, then groups code into 25 semantic workflows like Authentication or Billing. The 17 MCP tools let agents analyze PRs, detect drift via SHA-256 fingerprints, generate adversarial probes per workflow kind, allocate verification budgets with knapsack math, simulate rollbacks, and report execution results back so confidence scores decay realistically. Everything runs local, writes to SQLite, and exports an interactive HTML dashboard. No telemetry, no cloud tier. MIT licensed.