Pipecat is a Python framework for building real-time voice and multimodal conversational agents. It orchestrates audio and video streams, AI services, transports, and conversation logic as composable pipelines, so a single voice assistant or a multi-agent system can be assembled from modular processors.
Each pipeline is an agent. Pipelines can hand off to specialists, fan out in parallel, run as sidecar workers over a shared bus, or be distributed across processes and machines. Real-time interaction runs over transports such as WebSockets or WebRTC. Pipecat Flows, built into the framework, adds predefined or dynamic conversation paths with state management.
The surrounding tooling includes client SDKs for JavaScript, React, React Native, Swift, Kotlin, C++, and ESP32; a CLI (pipecat init) that scaffolds a project and can hand it to a coding assistant such as Claude Code or Codex, then monitors and deploys the agent; Whisker, a real-time debugger; Tail, a terminal dashboard; a Voice UI Kit of components and hooks; and Claude Code skills for scaffolding and deploying to Pipecat Cloud.
| Service category | Examples |
|---|---|
| Speech-to-text | AssemblyAI, AWS, Azure, Cartesia, Deepgram, ElevenLabs, Gladia, Google, Groq (Whisper), Mistral, NVIDIA, OpenAI (Whisper), Speechmatics, Whisper, xAI |
| LLMs | Anthropic, AWS, Azure, Baseten, Cerebras, DeepSeek, Fireworks AI, Gemini, Grok, Groq, and others |
| Other categories | Text-to-speech, speech-to-speech, transport, serializers, video, memory, audio processing, community integrations |
Features
- Composable pipelines: builds agent behaviour from modular processors for audio, video, and AI services
- Multi-agent composition: handoff, parallel fan-out, sidecar workers, and distributed deployment over a shared bus
- Real-time transports: streaming interaction over WebSockets or WebRTC
- Pipecat Flows: structured conversations with predefined or dynamic paths and state management
- Client SDKs: JavaScript, React, React Native, Swift, Kotlin, C++, and ESP32
- CLI: pipecat init scaffolds a project, and the CLI monitors and deploys agents to production
- Debugging and monitoring: Whisker for real-time pipeline debugging and Tail for a terminal dashboard
- Pluggable services: speech-to-text, LLM, text-to-speech, speech-to-speech, video, memory, and audio-processing providers, plus community integrations
- Coding-agent skills: Claude Code skills for project scaffolding and deployment to Pipecat Cloud
Integrates with
-
Cartesia
Real-time speech-to-text and text-to-speech via Cartesia's streaming APIs.
-
Daily
WebRTC transport for real-time audio/video agent communication.
-
Deepgram
Real-time speech-to-text (Nova/Flux) and Aura text-to-speech.
-
ElevenLabs
Expressive text-to-speech with word-level timing plus file-based transcription.
Integrated by
-
Async
Text-to-speech via Async's WebSocket and HTTP APIs.
Finchvox
Voice-AI observability that unifies conversation audio, logs, traces, and metrics for Pipecat agents.
Moss
Knowledge-retrieval service that injects semantic search results into an agent's LLM context.
-
Roark
Call analytics and observability for voice agents — lifecycle, transcripts, tool calls, and recordings.
