Multi-agent collaboration
Sitular
Build a room of specialists around the problem.
- Role
- Product, design & engineering
- Type
- Multi-agent workspace
- Status
- Private prototype
- Privacy
- BYO-key, browser-only
The friction
One generalist chat gives you one unexplained answer, unchallenged, with none of the trade-offs shown. Ad-hoc multi-agent setups are opaque, hard to steer mid-run, and they leak your API key to a server.
Where AI helps
Sitular assembles distinct specialists — each with a specialty and an operating lens — who respond to one another before an orchestrator reconciles consensus, disagreement, evidence, and next actions. You see the disagreement, you can pause and steer mid-run, and your key never leaves the browser.
Who it helps — Knowledge workers doing strategy and decision work — product, GTM, research, risk — who want structured, multi-perspective analysis instead of one chat answer.
What I built
-
A configurable agent roster
6 defaults or an AI-generated 4–6, each with one of three lenses: evidence, executive, or challenger.
-
Two run modes
Autonomous one-shot, or guided with pause-for-guidance checkpoints — plus per-run file and web grounding.
-
A bring-your-own-key architecture
Your key lives only in browser memory, never on disk or a server.
Proof