Pipecat

Pipecat is an open-source (BSD-2) Python framework for building real-time voice and multimodal AI agents by chaining speech-to-text, LLM and text-to-speech services into a low-latency pipeline. It is for engineering teams that want full control over their voice agent stack and the freedom to swap any model provider, with Pipecat Cloud available for hosting.

Pipecat is an open-source framework for voice agents, multimodal apps and real-time AI, built and maintained by Daily, the WebRTC infrastructure company, under the BSD-2-Clause licence. A Pipecat app has a client (browser, mobile app or phone) and a Python server that runs a pipeline of processors: audio comes in over a transport, goes through speech-to-text, an LLM and text-to-speech, and goes back out, with the framework managing the streams and keeping latency down. Every stage can be swapped. The GitHub repository lists dozens of STT, LLM and TTS providers, including Deepgram, AssemblyAI, OpenAI, Anthropic, Gemini, ElevenLabs and Cartesia, plus speech-to-speech services such as OpenAI Realtime and AWS Nova Sonic. Transports include Daily WebRTC, LiveKit, WebSockets, Vonage and WhatsApp. Client SDKs cover JavaScript, React, React Native, iOS, Android, C++ and ESP32. Smart Turn is an open-source end-of-turn model that uses cues like intonation, and Pipecat Flows handles structured, branching conversations. Teams can self-host, run on Pipecat Cloud (Daily-hosted, generally available since January 2026 according to Cekura), or use Pipecat Cloud Enterprise inside their own VPC. The repository had about 16.9k stars and 3.0k forks when checked on October 30, 2026. On October 29, 2026, Deepgram announced it is acquiring Daily for an undisclosed sum, and TechCrunch reported that Deepgram's CEO committed to keeping Pipecat open source and vendor-neutral. Its closest rival is LiveKit Agents; comparisons describe Pipecat as the more flexible, build-it-yourself option.

Category

Audio & Voice

Pricing

Free, Usage-based, Contact for pricing

Target Market

CTOs, Enterprise Developers, Heads of Customer Experience, AI Engineers

Deployment

Open-source, Self-hosted, Cloud-first

• ✓
  Composable pipelines

  Processors for transport, STT, LLM and TTS can be inserted, reordered or swapped without rewriting the rest of the agent.

• ✓
  Broad provider coverage

  Integrations with dozens of STT, LLM, TTS and speech-to-speech services, from OpenAI and Anthropic to Deepgram, ElevenLabs and NVIDIA.

• ✓
  Smart Turn detection

  An open-source, ML-based end-of-turn model that reads cues like intonation, so the agent interrupts users less often than silence-only detection.

• ✓
  Pipecat Flows

  Structured, state-based conversation flows for scripted tasks such as intake forms, qualification calls and multi-step support journeys.

• ✓
  Multiple transports and client SDKs

  Runs over Daily WebRTC, LiveKit, WebSockets, Vonage or WhatsApp, with clients for web, React Native, iOS, Android, C++ and ESP32.

• ✓
  Pipecat Cloud

  Managed hosting with per-minute agent profiles, unlimited concurrency, telephony, Krisp noise filtering and recording, so teams skip running their own fleet.

• •
  Inbound phone support agent

  Answer customer calls over PSTN or SIP, look up account data with LLM function calls, and transfer to a human with SIP REFER when needed.

• •
  Telehealth intake

  Run a voice intake conversation for patients on Pipecat Cloud with the HIPAA add-on and a signed BAA from Daily.

• •
  Provider bake-off

  Swap STT, LLM and TTS vendors inside the same pipeline to compare latency, accuracy and cost per minute before committing to a contract.

• •
  Outbound scheduling calls

  Dial out to customers to confirm or reschedule appointments, using Pipecat Flows to keep the conversation on a fixed sequence of steps.

• •
  On-device voice assistant

  Connect low-cost ESP32 hardware to a Pipecat server over WebRTC to build a voice assistant for kiosks, badges or devices.

The framework is free. Pipecat Cloud bills per agent minute by size ($0.01, $0.02 or $0.03 per active minute; reserved always-on agents $0.0005 to $0.0015 per minute), plus transport, telephony ($0.018/min PSTN) and recording. STT, LLM and TTS are billed by each provider and are usually the larger part of the bill. SSO, data residency, VPC deployment and volume rates are Enterprise, priced through sales. Rates read from daily.co on October 30, 2026.

This page was written from 8 sources, 7 on domains other than docs.pipecat.ai.

1. 1.github.com — pipecat
2. 2.docs.pipecat.ai — introductionvendor
3. 3.daily.co — pipecat cloud
4. 4.daily.co — security
5. 5.techcrunch.com — deepgram acquires pipecat maker daily to bolster its voice a
6. 6.cekura.ai — livekit alternatives
7. 7.channel.tel — pipecat vs livekit voice framework decision
8. 8.news.ycombinator.com — item