Your agent ran for 2 hours overnight. It edited files, ran commands, made autonomous decisions. You need to know what happened, not just the final diff.
## Agent session: "Refactor auth middleware" Duration: 1h 47m | 156 messages | 23 tool calls ### Files modified (12) - src/middleware/auth.ts: rewrote JWT validation (was 340 lines → 180) - src/middleware/auth.test.ts: added 14 new tests - src/routes/protected.ts: updated middleware reference - src/types/auth.ts: new type definitions ... and 8 more ### Commands run - npm test (3 times; first two had failures, third passed) - npm run lint (passed) - git diff --stat (agent checked its own changes) ### Key decisions 1. Split auth.ts into auth.ts + token.ts (separation of concerns) 2. Chose jose library over jsonwebtoken (better TypeScript support) 3. Added rate limiting to /auth/refresh (wasn't in original spec) ### Review needed - The agent added rate limiting that wasn't requested - New dependency: jose@5.2.0 (replaces jsonwebtoken) - 3 files deleted: old test fixtures
Agents are powerful but opaque. Without an audit trail, you're reviewing diffs blind. With Memo, you see the reasoning behind every change.