Momentic logs 117010 bugs caught across 70607819 test runs in production

Momentic supplies an agentic QA system that ingests plain-English YAML flows and executes them against live applications. The platform maintains a knowledge base of product terminology and expected behavior, then re-identifies UI controls at runtime when selectors change. Production telemetry shows 8932104 auto-heals and 339166 changes verified, indicating that runtime recovery plus reviewed repairs now replace most manual test maintenance.

Data from Notion, Coframe, Runway, and Webflow deployments confirm the pattern. Notion runs the suite on every pull request with 6.3x parallel execution; Coframe reports a measurable drop in post-merge regressions. These outcomes align with prior self-healing test literature (Memon 2008, Grechanik 2010) but add agent-driven triage that classifies failures as application bugs versus flakiness before surfacing evidence.

Operationally, the CLI exit codes integrate directly into existing CI gates, shifting verification from periodic regression suites to per-PR checks. This removes the historical trade-off between test coverage and release velocity for teams already using coding agents.

Next milestones will be measured by the fraction of tests that require no human review after the first heal and by adoption inside agentic coding loops that call Momentic via MCP.