LiveKit Agents
What is LiveKit Agents?
LiveKit Agents is a realtime AI agent platform for developers that builds, deploys, and monitors voice, video, and physical agents. It combines open source SDKs with Agent deployment, Agent observability, and turn-by-turn telemetry, plus Native client SDKs and Coding agent resources. LiveKit says it is used by OpenAI, Oracle, Salesforce, and Deutsche Telekom. Plans run Build $0/mo, Ship $50/mo, Scale $500/mo, and Enterprise custom.
Last verifiedHow we evaluate
At a glance
- LiveKit Agents is best for developers who need to build and monitor realtime voice or video agents at scale.
- Build $0/mo; Ship $50/mo; Scale $500/mo; Enterprise Custom
What does LiveKit Agents do?
LiveKit Agents handles the full lifecycle of realtime AI agents: you build with open source SDKs, deploy to LiveKit Cloud, and watch production behavior through realtime observability and session analytics. The platform routes media over WebRTC and HTTP/WebSocket paths, while features like Agent deployment, Agent observability, and turn-by-turn telemetry keep the runtime and debugging loop tight. Native client SDKs and coding-agent resources help teams move from prototype to production without changing the core agent code path. At scale, LiveKit reports 300,000+ developers, billions of calls annually, and 300+ AI model integrations. It supports voice, video, and physical AI agents, with LiveKit Inference covering 50+ models out of the box and another 200+ models available via plugins. The cloud layer adds global edge infrastructure, telephony, and metrics export APIs, while customers such as OpenAI, Oracle, Salesforce, and Deutsche Telekom signal production use across large teams.
Why use LiveKit Agents?
- Open source SDKs let teams keep agent logic in their own code while offloading production runtime management.
- Global edge infrastructure and autoscaling help reduce latency and absorb traffic spikes without manual capacity planning.
- LiveKit Inference plus plugin support gives teams broad model choice without rebuilding the agent stack.
- Observability, logs, and session reports make it easier to debug live conversations and review production behavior.
- No-code-change deployment to LiveKit Cloud lowers the friction of moving existing agents into production.
Who is LiveKit Agents for?
- Product engineers who need to ship realtime voice or video agents without managing runtime infrastructure.
- Platform teams who want deployment, observability, and scaling in one cloud workflow.
- AI developers who need flexible model routing across STT, LLM, and TTS providers.
- Teams building telephony-connected agents who need inbound calling and session analytics.
- Organizations with compliance needs who want region pinning and security reporting.
What are LiveKit Agents's key features?
Conversation handling
Manage live voice and text conversations with built-in telephony and one free US local phone number for inbound calling, so agents can answer real users immediately.
Native client SDKs
Use native SDKs to connect agents across apps and services, reducing glue code and making it easier to ship production workflows with LiveKit Agents.
Coding agent resources
Build coding assistants with resources for Claude Code, Cursor, and Codex, plus model access from OpenAI and Gemini for developer-focused agent workflows.
Global cloud network
Run agents on a global edge network with region pinning support, helping teams keep latency predictable and place workloads closer to users.
Elastic scaling
Scale from early tests to billions of calls annually with agent deployment and inference infrastructure designed for changing traffic without reworking your stack.
SIP support
Connect agents to phone systems through SIP support and telephony, which lets teams route calls into AI workflows without custom telecom plumbing.
Realtime metrics
Track session metrics and analytics in real time, giving operators visibility into call quality, usage, and failures while agents are running.
Recording exports
Export recordings and metrics through APIs, making it easier to review sessions, share evidence, and feed agent data into internal tools.
What does LiveKit Agents integrate with?
- Deepgram
- OpenAI
- Groq
- Cerebras
- Perplexity
- Cartesia
- ElevenLabs
- Claude Code
- Cursor
- Codex
- Gemini
What are LiveKit Agents's use cases?
Voice agents for product engineers
Product engineers use LiveKit Agents to ship realtime voice or video agents without building runtime infrastructure from scratch. They rely on Conversation handling to manage turn-taking and Native client SDKs to embed the experience cleanly in their app, so they can launch faster and keep the interaction polished.
Telephony agents with analytics
Teams building telephony-connected agents use LiveKit Agents to handle inbound calling and review each session after the fact. With SIP support and Session reports, they can trace call outcomes, spot drop-offs, and improve the agent before the next rollout.
Model routing for AI developers
AI developers use LiveKit Agents to route work across STT, LLM, and TTS providers without rewriting their stack. Inference and Conversation quality help them compare model behavior in production and tune the experience toward clearer, more natural responses.
Deployment control for platform teams
Platform teams use LiveKit Agents to deploy and scale agents in one cloud workflow. Agent deployment and Elastic scaling let them roll out updates, absorb traffic spikes, and keep operations predictable as usage grows.
How does LiveKit Agents work?
- Connect your first agent workflow in Agent deployment, then choose the runtime path for voice or video. Use the Global cloud network to place sessions close to users and reduce latency.
- Wire in your model providers through Inference, selecting STT, LLM, or TTS options as needed. Add Conversation handling so the agent can manage turns and respond naturally.
- Attach telephony with SIP support when you need inbound calling, then verify call flow in Realtime metrics. Use Recording exports to review sessions and catch issues quickly.
- Monitor production behavior in Agent insights and Agent logs, then export Session reports for stakeholders. Keep improving with Testing before each release and Elastic scaling as traffic changes.
How much does LiveKit Agents cost?
Build
$0/mo- No credit card required
- Start with:
- Agent deployment
- Agent observability
- Inference credits
- Global edge network
- Telephony (1 free number)
- Session metrics and analytics
- Community support
Ship
$50/mo- Everything in Build, plus:
- Team collaboration
- Custom voices
- Instant rollback to a previous agent deployment
- Email support
Scale
$500/mo- Everything in Ship, plus:
- Role-based access
- Metrics export APIs
- Region pinning
- Security reports / HIPAA
- Inference discounts
Enterprise
Custom- Everything in Scale, plus:
- Volume pricing, including inference
- Slack
- SSO
- Support SLA
Frequently asked questions
What is LiveKit Agents?
LiveKit Agents is a realtime AI agent platform for developers that builds, deploys, and monitors voice, video, and physical agents. It combines open source SDKs with Agent deployment, Agent observability, and turn-by-turn telemetry, plus Native client SDKs and Coding agent resources. LiveKit says it is used by OpenAI, Oracle, Salesforce, and Deutsche Telekom. Plans run Build $0/mo, Ship $50/mo, Scale $500/mo, and Enterprise custom.
How much does LiveKit Agents cost? Is it free?
LiveKit Agents has a free plan, with paid tiers including Ship at $50/mo, Scale at $500/mo, Enterprise at Custom.
What is LiveKit Agents used for? Who is it for?
LiveKit Agents is used for Conversation handling, Native client SDKs, and Coding agent resources. It's built for Product engineers, Platform teams, and AI developers.
Does LiveKit Agents have an API and what does it integrate with?
LiveKit Agents doesn't publish a public API. It integrates with Deepgram, OpenAI, Groq, Google, Cerebras, and 7 more.
Editor's read
Check whether the Build plan's included telephony and observability are enough for your rollout, or whether you'll need Ship for team collaboration and instant rollback. If region pinning, security reports, or metrics export APIs are required, those sit on Scale.
Explore other Agent Tools & Integrations
