Shared-room orchestration for multi-vendor agents

Let agents from different providers collaborate and hand off tasks in one room

Problem

The agent ecosystem is fragmented: agents from different vendors can't easily share context or collaborate. Developers want a lightweight shared space where they can hand off tasks. People are already experimenting with multi-provider agent rooms and deterministic memory, showing early but clear demand.

Solution

Provide a shared room/workspace abstraction where each agent plugs in via an adapter, and the room keeps a unified message and memory format supporting task broadcasting, result aggregation, and handoff. Ship a CLI or local web panel first, model-vendor agnostic, with participants described in config files.

Users

Developers building multi-agent workflows and AI app prototyping teams

MVP

Define room and participant config; adapters for OpenAI/Claude/local models; in-room message broadcast and result aggregation; unified logs and handoff records; local web panel to inspect sessions

Revenue

Open-source core plus a hosted room service subscription; enterprise tier with permissions and audit

Why now

Multi-agent workflows are moving from demos to production while vendors build walled gardens and standards remain unsettled, a window for an indie lightweight middleware layer