Pipecat is an open source Python framework from Daily for building voice and multimodal AI agents. You connect speech, language, and vision services into a streaming pipeline, so your agent can reply while the user is still talking. It handles interruptions and turn detection for you. Self-host it for free, or deploy on Pipecat Cloud for managed scaling, telephony, and monitoring. It fits developers building voice agents for phone, web, and apps.

• Multimodal Orchestration: Coordinate voice, video, text, and images in a single pipeline with frame-level control. You decide exactly how data moves through your agent.
• 200+ Swappable Providers: Switch speech-to-text, LLM, text-to-speech, and vision services, usually with one line of code. You can test a new model without rewriting your app.
• Natural Conversation Handling: Built-in interruption handling, turn detection, and context management make conversations feel less robotic.
• Multi-Agent Workflows and Pipecat Flows: Hand off long-running tools to subagents, or use Pipecat Flows to build structured conversation paths for guided interactions.
• WebRTC, SIP, and PSTN Transports: Connect your agent to browsers, apps, and phone lines. Client SDKs and a UI kit help you build the front end.
• Evals and Observability: Run evals instead of manual test calls, and inspect each turn with metrics, traces, and OpenTelemetry support.
• Faster project setup: Use the CLI and starter templates to create projects without wiring every component from scratch.

• Free and open source: The core framework costs nothing, and you can run it anywhere Python runs, on your own infrastructure.
• Large integration library: More than 200 integrated providers let you pick the best speech, language, and vision models for your needs and swap them as the market changes.
• Deep control over the pipeline: You can customize each processing step, which comparisons say managed platforms often don't allow.
• Flexible deployment paths: Start self-hosted, move to Pipecat Cloud for managed scaling, or run Pipecat Enterprise inside your own AWS, Azure, or GCP environment.
• Usage-based cloud pricing: Pipecat Cloud offers agent runtime pricing from $0.01 per minute, with additional charges possible for AI services, telephony, and infrastructure.

• Code-first, Python-focused: Pipecat is a developer framework, not a visual builder. Non-technical teams will need engineering help, or a managed no-code voice platform.
• Self-hosting means operational work: The free option requires you to deploy, scale, and monitor the infrastructure yourself. Managed platforms take care of that for you.

• AI Phone Agents: Connect an agent to PSTN or SIP to answer inbound calls or place outbound ones. Your support or sales line can handle routine calls around the clock.
• Customer Support Voice Bots: Use Pipecat Flows to guide callers through structured paths such as account checks or troubleshooting. Customers get consistent answers and handoffs to the right tool.
• In-App Voice Assistants: Add a real-time voice interface to your web or mobile app using WebRTC and the client SDKs. Users can talk to your product instead of typing.
• Multimodal Agents: Build agents that see and hear, combining video, images, and speech in one pipeline. Useful for tutoring, visual inspection, or interactive demos.
• Model Benchmarking and Prototyping: Swap STT, LLM, and TTS providers with minimal code changes to compare latency, quality, and cost. Teams can pick their stack based on real tests rather than guesses.
• Regulated Enterprise Deployments: Run Pipecat Enterprise in your own cloud account with enterprise support. Teams in healthcare or finance can keep data in their environment while still using managed tooling.