Documentation
Impact Gate starts with deterministic diff analysis.
Review pull requests, release branches, hotfixes, and shipped tags. Inspect existing spec associations and their provenance before choosing which scenarios to author or run first.
Review → Inspect → Test
Build the current source, export a reviewable scenario plan, and check spec-mapping policy. A passing gate does not establish measured coverage or release safety.
How the evidence layer works
Learn route families, confidence scoring, plan artifacts, and where AI fits into the workflow without becoming the workflow.
See where it fits against commercial platforms
Compare Impact Gate honestly against hosted optimization, AI-first E2E, no-code automation, and monitoring-oriented tools.
Inspect the changes since the previous release
Compare the current candidate to the last shipped tag and turn that delta into review context while retaining the required regression suite.
Ground generation against what your repo already knows
Review generation uses local API context. Unaccepted proposals stay in quarantine; acceptance currently requires direct local-source mutation evidence. Browser application tests remain unverified by that verifier.
Explore one running user journey
Give the agent a running app URL and a bounded goal. Inspect the resulting findings and reproduction steps before authoring tests.
Layer generation and healing on after the plan is useful
Configure providers, inspect artifacts, and add autonomous workflows only after the deterministic path is already trustworthy.
CI, troubleshooting, and cost control
Use focused playbooks for PR gating, release diffs, AI safety, budget controls, and rollout troubleshooting.