Agent Swarm
The full 11-agent strategic team. Every important assignment gets 10 parallel thinkers, a cross-read pass, and an independent maker-checker before anything ships.
Run Archive
The system was further along than expected. The `/portal/[client]` page already had summary cards and a stage timeline. A separate page at `/deals/hr-com/meetings` already had 268
The Agent Swarm Mode feature system spans 107+ files across the monorepo: 53 files in the maxswarm skill alone, 20 in swarm-upgrade, 2 Next.js pages, 7 public HTML archives, 14 eng
There are no calls to analyze. Across all four data sources β Salesfinity, Fireflies, SpokePhone, and Gmail β zero call records were found for the period 2026-04-27 to 2026-05-04.
Consolidating saves ~$12k/year for a 1-engineer team, with a six-week payback. The plan works, but three high-risk steps (the Lob webhook re-pointing, the deals Supabase migration,
The swarm produced a complete 38-document data room organized into 8 sections, with every document containing a summary statement, a "Why This Matters" explanation, numbered fill-i
One signal was fully deployed. Three were not. The one that was fixed (caching confirmed sponsors in Supabase) was fixed today β in this session β only because Ewing directly asked
The deal-tracking system was eventually shipped in `next-chapter-os/src/app/room/[slug]/` as the deal-room route, with `v_deal_room_content` view and `deal_plays` table populated f
(filled in Phase 6)
How the swarm works
All 10 working agents get the same brief and produce independent answers. No agent sees another's draft yet.
Every agent reads all 9 peer drafts, then re-examines their own. They must find where they were wrong before defending their position.
Audit-quality sees only the synthesized final output β never the drafts. Returns PASS / FIX-AND-SHIP / REWORK. Conductor acts on the verdict.
Swarm runs are archived here and in ~/.claude/skills/maxswarm/previous_swarms/ on the Mac Mini.