Skip to main content

LiveKit Cloud

An end-to-end platform for building, deploying, and operating AI agent applications.

Overview

LiveKit Cloud is a fully managed, globally distributed platform for building, hosting, and operating AI agent applications at scale.

While LiveKit's open-source server provides the realtime media foundation, LiveKit Cloud extends beyond managed infrastructure. It combines realtime audio, video, and data streaming with agent development tools, managed agent hosting, built-in inference, native telephony, and production-grade observability in a single, cohesive platform.

What LiveKit Cloud includes

Realtime communication core: A fully managed, globally distributed mesh of LiveKit servers that powers low-latency audio, video, and data streaming for realtime applications.

Agent Builder: Design, test, and iterate on AI agents using a purpose-built development experience. Agent Builder streamlines prompt design, tool configuration, and interaction flows.

Managed agent hosting: Deploy and run agents directly on LiveKit Cloud without managing servers or orchestration. LiveKit handles scaling, lifecycle management, isolation, and upgrades.

Built-in inference: LiveKit Inference lets you run supported AI models directly within the LiveKit Cloud environment without requiring API keys. It's zero data retention by default, so prompts, audio, and model outputs are never stored or used to train models.

Native telephony: LiveKit Phone Numbers lets you provision phone numbers and connect PSTN calls directly into LiveKit rooms without setting up trunks.

Observability and operations: Production-grade analytics, logs, and quality metrics are built into the LiveKit Cloud dashboard, giving visibility into agent behavior, media quality, usage, and performance across your deployment.

Why choose LiveKit Cloud?

  • End-to-end platform: Build, deploy, and operate AI agents, realtime media, inference, and telephony in one system.
  • Zero operational overhead: No need to manage servers, scaling, or infrastructure.
  • Global edge network: Users connect to the closest region for minimal latency.
  • Elastic, unlimited scale: Support for rooms with unlimited participants using LiveKit's global mesh architecture.
  • Enterprise-grade reliability: 99.99% uptime guarantee with redundant infrastructure.
  • Comprehensive analytics: Monitor usage, performance, and quality metrics through the LiveKit Cloud dashboard.
  • Seamless developer experience: Use the same APIs and SDKs as open source, with additional cloud-native capabilities.

Open source compatible, platform complete

LiveKit Cloud runs the same open-source LiveKit server available on GitHub  and supports the same APIs and SDKs. This means:

  • You can start on open source and migrate to LiveKit Cloud without rewriting application code.

  • You can move from LiveKit Cloud to self-hosted if your requirements change.

  • Your client and agent code remains portable — the connection endpoint is the primary difference.

What does differ is everything around the server: agent tooling, hosting, inference, telephony, global scaling, and observability, all of which are native features of LiveKit Cloud.

Comparing LiveKit Cloud to self-hosted

When building with LiveKit, you can run the open-source server yourself or use LiveKit Cloud as a fully managed, end-to-end platform.

AI agents

Self-hostedLiveKit Cloud
Agents frameworkFull supportFull support
Managed agent hostingN/A, you run and scale agent servers yourself.Included: deploy agents to LiveKit Cloud with automatic scaling, managed builds, and instant rollback.
Agent BuilderN/AIncluded
Built-in inferenceN/A, bring your own model provider API keys.Included, with custom voice clones on paid plans.
Agent observabilityCustom / external, using OpenTelemetry data hooks.Included: built-in observability stack with transcripts, traces, session recordings, and log drains.
Noise cancellationClient-side WebRTC cancellation; ai-coustics supported with your own license key.Client-side WebRTC cancellation, plus enhanced models (Krisp, ai-coustics).

Telephony

Self-hostedLiveKit Cloud
SIP & telephonyFull support, SIP service deployed as a separate service.Full support
Phone numbersBring your own SIP trunking provider.Native LiveKit Phone Numbers, plus support for external SIP trunks.
Noise cancellationN/AKrisp noise cancellation applied at the SIP trunk for inbound and outbound calls.

Media and transport

Self-hostedLiveKit Cloud
Realtime media (audio, video, data)Full supportFull support
Egress (recording, streaming)Full support, deployed as a separate service.Full support, included with every project.
Ingress (RTMP, WHIP, SRT ingest)Full support, deployed as a separate service.Full support, included with every project.
End-to-end encryptionFull supportFull support
WebhooksFull supportFull support

Platform and operations

Self-hostedLiveKit Cloud
Who manages itYouLiveKit
ArchitectureSingle-home SFUGlobal mesh SFU
Connection modelSingle server per roomEach user connects to the nearest edge.
Max users per roomUp to ~3,000No limit
Regions & data residencyRegions you deploy toGlobal edge network, with region pinning for data residency requirements.
Analytics & telemetryCustom / externalLiveKit Cloud dashboard, plus Analytics API on Scale plans.
Access token revocationN/AIncluded
Uptime guaranteesN/A99.99%

LiveKit Cloud administration

For information about LiveKit Cloud architecture, administration, and configuration, see the Administration section.

Next steps

Ready to deploy your agents? Get started with the Agent deployment guide.