LiveKit Agents

LiveKit Agents is an open-source framework and cloud platform for building voice, video, and physical AI agents. Python or Node.js programs can join LiveKit rooms as realtime participants, and deployments can run on LiveKit Cloud or a developer-selected environment. Agents can process realtime voice, video, and text input and produce output in any modality. The framework supports LLM tool use, frontend RPC calls, and multi-agent handoffs, as well as phone connections through native SIP. LiveKit Inference provides models from providers including OpenAI, Google, AssemblyAI, Deepgram, Cartesia, and ElevenLabs; open-source plugins connect to additional providers. LiveKit Cloud offers session replay, transcripts, trace spans, logs, and session reports. The free Build plan includes 1,000 agent session minutes, five concurrent sessions, and one deployment. Ship starts at $50/mo and includes 5,000 session minutes, then $0.01 per minute. The Agent Builder is intended for simple voice-agent prototypes and produces Python code for deployment to LiveKit Cloud.

LiveKit Agents combines realtime agent development with deployment choices, provider integrations, and phone connectivity. The free Build tier has defined session and deployment limits, while Ship adds capacity starting at $50/mo.

LiveKit Agents suits developers building realtime voice, video, or physical AI agents who want Python or Node.js SDKs and a choice of cloud or custom deployment. Its phone integration and observability features may also suit teams building agents for calls and monitoring sessions.

Overview

LiveKit Agents is an open-source framework paired with a cloud platform for building realtime voice, video, and physical AI agents. It is best suited to developers who want to build in Python or Node.js and retain a choice of managed or custom deployment. Its standout advantage is breadth across media, models, workflows, and telephony; the tradeoff is that it is a framework for building agents, not a finished assistant to configure.

Key features

Python and Node.js programs can join LiveKit rooms as realtime participants, handling voice, video, and text input and producing output in any modality. That makes LiveKit Agents a stronger fit for interactive, live-media experiences than for teams seeking a text-only assistant. LLM tool use, frontend RPC calls, and handoffs between agents support more involved workflows, but require a development team to put those capabilities to work.

Deployment can be on LiveKit Cloud or in a custom environment, giving teams a choice between managed hosting and their own setup. LiveKit Inference connects to models from OpenAI, Google, AssemblyAI, Deepgram, Cartesia, and ElevenLabs; open-source plugins extend the provider options. Native SIP connects agents with phone systems, and the platform supports inbound and outbound calls, custom phone numbers, and human handoff.

LiveKit Cloud's agent insights include session replay, transcripts, trace spans, logs, and session reports, useful for inspecting agent sessions. LiveKit says prompts, audio, and completions handled by LiveKit Inference and its model providers are not logged or retained. The company lists SOC 2 Type II, GDPR, HIPAA, CCPA / CPRA, and the EU–US Data Privacy Framework; PCI DSS and ISO 27001 are in progress. BAAs are available to Scale and Enterprise customers handling protected health information.

Agent Builder is aimed at simple voice-agent prototypes and produces Python code for deployment to LiveKit Cloud. It is a quick starting point for that narrow use, but not a substitute for the framework when a project needs a custom environment or more complex design.

Pricing

Build costs $0/mo and includes 1,000 agent session minutes, 5 concurrent agent sessions, 1 agent deployment, $2.50 in inference credits, and community support. It is a practical way to start a small project, but the session and deployment caps make it a limited option for sustained or concurrent workloads.

Ship starts at $50/mo, with 5,000 agent session minutes, then $0.01 per minute, 20 concurrent sessions, 2 deployments, and email support. This is the entry paid tier for teams that need more headroom than Build provides; the per-minute charge matters once included minutes run out.

Scale starts at $500/mo and includes 50,000 session minutes, then $0.01 per minute, up to 600 concurrent sessions, 4 deployments, region pinning, security reports, and HIPAA. Its capacity and compliance-related features are aimed at larger or more demanding operations, at a substantially higher starting price. Enterprise has custom pricing and adds volume pricing, a shared Slack channel, SSO, and a support SLA. Build has community support, Ship email support, and only Enterprise promises a support SLA.

Platforms

LiveKit Agents supports Android, iOS, Linux, macOS, Windows, web, API, and self-hosted environments. That range accommodates both app-facing projects and teams deploying their own infrastructure.

Who it's for

Choose LiveKit Agents if you have developers building realtime agents and value control over deployment, integrations, and phone connectivity. It is less suitable if you want a ready-made assistant without framework work, or if the Build tier's session and concurrency caps are too restrictive for your use.

Pros and cons

Pros

• Realtime voice, video, and text support, plus outputs in any modality, suits agents built for live interaction.
• Cloud or custom deployment, multiple model-provider connections, and open-source plugins leave teams room to shape their stack.
• Native SIP and support for inbound and outbound calls, custom numbers, and human handoff cover phone-based workflows.
• Cloud session replay, transcripts, trace spans, logs, and reports offer several ways to inspect agent sessions.

Cons

• It is a developer framework, so teams seeking a finished assistant rather than a build platform should look elsewhere.
• Build is capped at 1,000 session minutes, 5 concurrent sessions, and 1 deployment.
• Ship charges $0.01 per minute after its 5,000 included minutes, so usage above that allowance adds cost.
• HIPAA BAAs are available only to Scale and Enterprise customers, not to teams on the lower tiers.

Alternatives

AI Voice Agents and Conversational AI Platforms are useful starting points for comparing options across those categories. Consider Yellow.ai if its free plan's AI-agent and campaign allowances suit your needs. Pipecat is another free option. Choose Tone if you want its free self-hosted plan with unlimited agents, calls, and workflows and bring-your-own model, voice, and telephony. Rapida AI offers a free Launch plan with one agent and bring-your-own STT, TTS, LLM, and SIP. Ring AI has a free tier with 100 minutes per month for testing. Deepgram Voice Agent API is a paid alternative with pay-as-you-go Standard and bring-your-own-TTS plans. Cognigy is another paid web option. Intervo offers a freemium option with self-hosted open-source code under an MIT License.

Verdict

LiveKit Agents is a strong choice for developer teams building realtime voice, video, or phone agents that need deployment flexibility and broad integration options. Its main reason to choose it is the combination of live-media workflows, cloud observability, and native SIP in one framework. Look elsewhere if you need a ready-to-use assistant or if your workload quickly outgrows the free tier and the paid starting prices do not fit.