tools
LiveKit Agents
Open-source framework for building, deploying, and observing realtime voice, video, and physical AI agents.
LiveKit Agents lets developers build realtime agents with Python or Node.js. Agents can process voice, video, text, and data, use tools and multiple AI providers, and connect to web, mobile, or phone-based users.
Teams can run agents locally or on their own infrastructure, or deploy them to LiveKit Cloud for managed hosting, scaling, telephony, observability, and analytics. Cloud usage beyond included quotas, model inference, telephony, and other resources may cost extra.
Features
- Build realtime voice, video, and text agents with Python or Node.js
- Connect LLM, STT, TTS, realtime API, and avatar providers
- Support tool use, MCP, RPCs, and data APIs
- Coordinate multi-agent handoffs and shared workflows
- Deploy with managed jobs, dispatch, scaling, and load balancing
- Connect agents to inbound and outbound phone calls through SIP
- Monitor sessions with transcripts, traces, metrics, and recordings
- Apache-2.0 licensed open-source Agents framework
Use cases
- Build customer-service call center agents
- Create multimodal assistants for voice, text, and screen sharing
- Support realtime telehealth consultations
- Translate conversations in realtime
- Add language-model-backed NPCs to interactive applications
- Connect cloud-based agents to robots
Pros
Cons
Latest updates
- livekit-agents@1.8.11 (1.8.11)
Responses API is available through LiveKit Inference; added Zoom Scribe, Kalpa TTS, Rime STT and Prism, and Fish Audio transcribe-stream.
- livekit-agents@1.8.10 (1.8.10)
RecorderIO adds bitrate and encoding-mode controls; Reson8, Mistral Voxtral 3, and Inworld models are added.
- livekit-agents@1.8.9 (1.8.9)
LiveKit Inference adds named decision evaluations; custom LLM nodes can attach metadata to provider request spans.
- livekit-agents@1.8.8 (1.8.8)
GPT-Live becomes available through LiveKit Inference, with added Gemini Live model strings and updated Grok Voice models.
- livekit-agents@1.8.7 (1.8.7)
Realtime assistant replies expose end-to-end latency metrics; Claude Opus 5, Opus 5.5, and Sonnet 5.5 are supported.
Capabilities
- Command line — “deploy to LiveKit Cloud using the CLI or with a single click in the dashboard.” source
- Choice of models — “the LiveKit Agents framework supports another 200+ models via plugins” source
- Self-hosted — “Self-hosted agents get a development allowance, then fall back to local models.” source
- No code retained — “Prompts, audio, and model outputs aren't logged or stored in LiveKit or underlying model providers” source
- API — “Simple and Powerful APIs” source
- Official SDKs — “Build voice agents with open source SDKs and complete developer control.” source
- Runs models for you — “Access STT, LLM, and TTS models with optimized and low-latency routing.” source
- Builds agents and workflows — “An open source framework and developer platform for building, testing, deploying, scaling, and observing agents in production.” source
- Traces and evaluates — “Replay sessions, review transcripts, and inspect trace spans.” source
Security
- SOC 2 Type II — “SOC 2 Type II” source
- SOC 2 — “SOC 2 Type II” source
- Not: ISO 27001 — “PCI DSS and ISO 27001 are in progress.” source
- GDPR — “DPA available; sub-processor list maintained publicly.” source
- HIPAA — “BAAs available for Scale and Enterprise customers handling protected health information.” source
Pricing
- Starting price
- $50/mo
- Prices checked
- 2026-10-25
Build
Free
- Agent deployment
- Agent observability
- Agent simulations
- Inference credits
- Global edge network
- Telephony (1 free number)
Ship
- STARTING AT $50/mo
- Team collaboration
- Custom voices
- Instant rollback to a previous agent deployment
- Email support
Scale
- STARTING AT $500/mo
- Role-based access
- Metrics export APIs
- Region pinning
- Private links
- Security reports / HIPAA
- Inference discounts
Enterprise
Price on request
- Volume pricing, including inference
- Shared Slack channel
- SSO
- Support SLA