Pipecat

Pipecat is an open-source Python framework for building real-time voice and multimodal conversational agents, maintained by Daily. You compose STT, LLM, and TTS services into a frame-based pipeline — audio flows in, text reasoning happens in the middle, speech flows back out — with the framework handling voice activity detection, turn-taking, interruptions (barge-in), and function calling. It supports 50+ AI service integrations, multiple transports (WebRTC via Daily or LiveKit, WebSocket, WhatsApp, Twilio), and multi-agent patterns (handoff, parallel fan-out, sidecar workers) that can run locally or distributed across machines.

Quickest path — the CLI scaffolds a runnable bot in under a minute (README-verified):

```bash

uv tool install "pipecat-ai[cli]"

pipecat init quickstart

```

Or add to an existing project:

```bash

pip install pipecat-ai

```

A minimal voice assistant pipeline looks like this:

```python

from pipecat.pipeline.pipeline import Pipeline

from pipecat.pipeline.runner import PipelineRunner

from pipecat.pipeline.task import PipelineTask

pipeline = Pipeline([

transport.input(), # listen to the user

stt, # speech -> text (e.g. Deepgram)

llm, # reason + respond (e.g. OpenAI)

tts, # text -> speech (e.g. Cartesia)

transport.output(), # send audio back

])

await PipelineRunner().run(PipelineTask(pipeline))

```

Full quickstart: https://docs.pipecat.ai/getting-started/quickstart