Developer Preview
Agents
Build programmable participants that can join sessions and process or synthesize media streams in real-time.
Compose multimodal AI apps with plugins
Chain together prebuilt integrations for speech and video processing, text or image generation, and more.
Build on top of our growing set of plugins
Use WebRTC on the server
Your agent has access the same LiveKit WebRTC features as our client SDKs.
1async def publish_audio(self):2 audio_source = rtc.AudioSource(sample_rate, num_channels)3 track = rtc.LocalAudioTrack4 .create_audio_track("agent-voice", audio_source)56 options = rtc.TrackPublishOptions()7 options.source = rtc.TrackSource.SOURCE_MICROPHONE89 await self.ctx.room.local_participant10 .publish_track(track, options)
1
Create agents that can see and hear
Pipe streaming, raw audio packets or video frames through an AI model or processing workflow.
1async def detect_faces(self, frame: rtc.VideoFrame):2 faces = detect_faces(frame)3 annotated_frame = draw_bounding_boxes(faces, frame)4 video_source.capture_frame(annotated_frame)
1
Example showing detecting a face in realtime
Stream output from generative AI
Generate and stream voice or video samples on-the-fly or play back prerecorded media files.
1async def synthesize(self, text: str):2 generated_frame = await stable_diffusion_generate(prompt=text)3 video_source.capture_frame(generated_frame)
1
Autoscale and orchestrate your agents in prod
LiveKit’s edge worker service automatically monitors, health checks, and distributes load across your pool of agents.
Allow machines to see the world
DirectAI works with LiveKit Agents to help developers instantly build computer vision models without code or training data.
See agents in action
Browse some of our agent demos that leverage the latest in AI powered technology
Try Agents for free
LiveKit Cloud is a cloud real-time platform and the fastest way to prototype and ship to production.