37×
fewer tokens across 15 codebases
Fifteen codebases, one session, two minutes of discovery
Cross-project questions — which app has the best contact module, are they all on the same Next.js, what tables every project shares — answered by graphing 15 production codebases at once: ~100,000 nodes, ~300,000 edges, built in ~30 seconds, queried in a single session.
Done with Explore agents, this is 120 agent spawns across 5–10 sessions — ~2.175M tokens, more than any one context holds, losing the cross-project comparison every time the window resets. The graph indexed all 15 repos in parallel in ~30 seconds and kept every result in one session.
That made intersection queries possible that agents can’t do at all: 18 foundation tables present in every project, entity-by-entity “which app has the most complete implementation,” and dependency drift — 9 of 10 still on TypeScript 5, Supabase 12 minor versions behind in 5 repos. ~58,500 tokens against ~2.175M, and ~2 minutes against an estimated ~3 hours of manual work.
- Cross-project intersection queries are impossible once agents force a context reset
- ~100,000 nodes and ~300,000 edges indexed in ~30 seconds, in parallel
- Foundation tables, entity completeness, and version drift — all data, not guesswork
This is what Pro delivers.
Not features for their own sake — measurable leverage on every session.