Pipecat is an open-source Python framework for building voice and multimodal AI agents that talk in real time. Developers compose a pipeline in which frames stream through speech recognition, language model and speech synthesis stages, so an agent can begin responding while the user is still speaking. It coordinates voice, video, text and images in one pipeline, handles interruptions and turn detection, and lets you swap any of more than 200 integrated providers, usually with one line of code. The framework is free and self-hostable, and it is maintained by Daily with support from its developer community.
Pipecat is for software engineers and AI teams who want control over how a voice agent is built rather than a closed no-code product. Startups prototype conversational bots with it, and enterprises use it for production agents, with Pipecat Cloud offering managed hosting and Pipecat Enterprise offering deployment in your own AWS, Azure or GCP environment with enterprise support. Teams new to the project can run the pipecat init command to scaffold a working project, which also connects a coding agent to current Pipecat documentation through an MCP server.
Typical projects include phone agents that connect over SIP or PSTN, browser-based voice assistants using WebRTC, and customer service bots that hand off to subagents for long-running tasks. Developers add tools so an agent can look up data or take actions, use Pipecat Flows to keep a conversation on a defined path, and run evals instead of manual test calls. Client SDKs and a UI kit help build front ends, while metrics, traces and OpenTelemetry support show what happened in every turn of a conversation.
