AI agent regression memory - Anchor Docs

Retrieve local regression memory from PRs, reverts, hotfixes, root-cause notes, and cross-repo impact before AI agents edit risky code.

Agents can make plausible edits that repeat an old incident, rejected approach, or production regression if that history is not in context.

• Extracts deterministic regression signals from PR history, labels, comments, reviews, commits, and affected files.
• Ranks regression memory higher when files, symbols, API contracts, or org consumers match the current task.
• Supports strict mode so weak or stale regression evidence is filtered out for high-risk changes.

anchor review --strict
anchor explain src/api/access.ts
anchor org impact --org my-org --strict

Anchor docs · GitHub