Pipecat
Open-source Python framework for voice and multimodal AI agents with swappable STT, LLM and TTS services, plus hosted Pipecat Cloud
Overview
Pipecat is an open-source (BSD-2-Clause) Python framework for building voice and multimodal AI agents as pipelines: audio comes in, flows through speech-to-text, an LLM and text-to-speech (or a speech-to-speech model), and goes back out in real time. Its site calls it "the most widely used open source ecosystem" for voice agents, with 200+ provider integrations, plus companion projects such as Pipecat Flows for structured conversations, the Smart Turn turn-detection model and the SmallWebRTC transport. Pipecat was created by Daily, which also runs Pipecat Cloud for hosting agents. On 29 October 2026 Deepgram announced it is acquiring Daily; Deepgram says Pipecat stays BSD-licensed and is moving to a Pipecat Foundation with a steering committee that includes other companies.
The Verdict
Who Should Use Pipecat?
Best For
- Python developers who want full control over every step of a voice pipeline
- Teams that want to swap STT, LLM and TTS vendors freely
- Self-hosting a voice agent, or hosting it cheaply on Pipecat Cloud
Not Ideal For
- No-code teams (look at hosted voice agent platforms instead)
- Teams that want turn-taking and interruption handling tuned out of the box
- Buyers who need a neutral vendor guarantee before the Deepgram deal settles
What's Great
- Open source, with a very large set of plug-in services
- Per-step control of the pipeline (frames, processors, transports)
- Pipecat Cloud is cheap to start: from $0.01 per active agent minute
- Used by large AI companies, according to logos on its site (NVIDIA, Anthropic, AWS)
Watch Out For
- Turn-taking and voice activity detection need tuning for natural calls
- Ownership is changing: Deepgram is acquiring its maker Daily, foundation still being set up
- Few independent reviews; most comparisons online are written by vendors
Pricing
View all features & details
Key Features
- Pipeline architecture for realtime audio, video and text
- 200+ integrated STT, LLM, TTS and vision services (vendor figure)
- Pipecat Flows for structured, multi-step conversations
- Smart Turn open model for turn detection
- SmallWebRTC and Daily WebRTC transports, telephony via Daily SIP and PSTN
Pipecat Cloud Extras
- Krisp noise cancellation: first 10k active session min/mo free, then $0.0015/min
- Daily WebRTC transport and recording billed per minute
Use Cases
- Phone and web voice agents
- Voice interfaces for existing LLM apps
- Multimodal agents that see and talk
How It Compares
| Feature | Pipecat | LiveKit | Daily | Vapi |
|---|---|---|---|---|
| What it is | Transport-agnostic agent framework | Media server plus agent framework | WebRTC API platform | Hosted voice agent API |
| Open source | Yes (BSD-2-Clause) | Yes (Apache-2.0) | No | No |
| Language | Python | Python, Node.js | SDKs for web and mobile | API and SDKs |
| Hosted option | Pipecat Cloud | LiveKit Cloud | Daily | Vapi |
| Entry price | Free OSS; Cloud from $0.01/min | Free Build plan | 10,000 free min/mo | $0.05/min platform fee |
Is this your tool? Add a badge