Opens in a new tab

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
Sitular autonomous workspace with a strategy brief and three specialist agents ready to collaborate

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

Gemini · BYO-key6 specialist typesAutonomous + guidedZero-dependency server