Observe and enrich
Transcription, translation, moderation, vision, analytics, and custom worker adapters receive participant and track lifecycle context and publish structured results back into the room.
One participant fabric for every live workflow.
Coordinate people, AI agents, robots, devices, applications, and backend services through shared identity, presence, media, state, commands, and operational context.
participantCustomer site
signalDevice alert
AI workerAssess + route
operatorAuthorized action
Human operatorsAI workersService participantsRobot + device bridgesInspectable workflows
Make AI part of the governed live workflow.
AI joins with identity, presence, permissions, outputs, and lifecycle instead of operating as disconnected webhooks beside the customer experience.
Render agent presence
Presence surfaces turn agent state, audio energy, and speech phases into local or publishable personas. Browser AI perception remains an optional plugin layer.
Speak locally
The local TTS foundation supports native browser speech and self-hosted Piper-WASM through a shared provider API, including publishable Web Audio tracks.
Know the boundary
A complete server-side conversational voice-agent runtime does not exist yet. Production turn-taking, barge-in, and STT→LLM→TTS orchestration remain roadmap work.
Resolve issues with people, equipment, and diagnostics in one session.
Experts, customers, connected equipment, diagnostics, and supporting services share the same controlled operational context.
Operator and customer rooms
Video, screen share, chat, files, role-aware controls, guest admission, and room short links for guided support sessions.
Diagnostics in the room
Surface ICE paths, peer state, quality changes, topology, timelines, device capabilities, and redacted support bundles while the session is active.
Bring the equipment
Use MQTT, serial, BLE, or USB gateways to make local equipment state and commands available to authorized room participants.
Run near the workflow
Deploy in a customer cloud, at the edge, or on private networks with explicit TURN, SFU, firewall, health, and capacity guidance.
Bring robotics and connected equipment into the product.
Robots and devices become governed participants that publish state, receive authorized commands, and share media with operators, agents, and services.
ROS2 robot participants
Native rclpy and rosbridge paths connect topics, services, state, and camera workflows to TomatoRTC rooms.
Browser teleoperation
Keyboard and gamepad controls map to command topics with deadman behavior, rate limits, acknowledgements, and operator HUDs.
Industrial gateway pattern
Thin MQTT, serial, BLE, and USB sidecars connect local hardware without embedding WebRTC or room logic in device firmware.
Digital-twin room state
Shared signals, awareness, and object foundations coordinate sensor data, commands, dashboards, simulations, and agent-visible state.
Honest deployment boundary: simulation-first demos and bridge packages are available, but broader robot hardware profiles and real-hardware integration matrices remain active engineering work.
Evaluate a live workflow with TomatoRTC.
Start with the business outcome, participants, control boundaries, media paths, and deployment constraints.