Livekit Transitions to AI Agent-Centric Pricing Model, Introducing Agent Session Minutes and Inference Credits

LiveKit restructured pricing from connection-based model (concurrent participants, connection minutes, bandwidth) to AI agent-centric model (agent session minutes, inference credits, agent deployments)

Primary usage metric changed from 'connection minutes' (time connected to network) to 'agent session minutes' (time in session with AI agent)

Agent session minutes metric introduced: Build 1,000 minutes, Ship 5,000 minutes, Scale 50,000 minutes

Inference service with access to STT, LLM, and TTS models via single API key added

Cold start prevention feature added for keeping agents always-on for instant responses

Agent deployments limit introduced: Build 1 deployment, Ship 2 deployments, Scale 4 deployments, Enterprise unlimited

Deployment metrics dashboard showing resource allocation, latency, errors, and process metrics added

Inference credits introduced: Build $2.50 credits (~50 min), Ship $5 credits (~100 min), Scale $50 credits (~1,000 min)

Agent deployment hosting on LiveKit Cloud infrastructure added as core feature

Instant rollback to previous agent deployment version added

Concurrent agent sessions limit introduced: Build 5 sessions, Ship 20 sessions, Scale up to 1,000 sessions