Introduction
Get Started
Piopiy is an open-source Python framework for building real-time voice AI agents that can listen, reason, and speak on telephony and WebRTC transports. It orchestrates AI services, transport layers, and call/session handling so you can focus on agent behavior instead of low-level voice plumbing.
Want to start immediately? Go to Quickstart.
What You Can Build
Voice Assistants
Natural spoken conversations for support, intake, and FAQ workflows.
Phone Agents
Inbound and outbound call agents connected directly to phone numbers.
Knowledge Agents
Search external databases and documents using Model Context Protocol (MCP).
Tool-Calling Agents
Call internal APIs and business tools during live conversations.
Dynamic Voice Flows
Switch providers or voices at runtime based on conversation context.
Call Transfer
Seamlessly hand off a live session to a human support agent.
How It Works
Piopiy powers real-time voice AI using the TeleCMITransport, acting as a high-speed bridge between telephony networks and your AI.
The Live Pipeline:
- Listen: TeleCMITransport receives the live call and streams audio to your STT (Speech-to-Text).
- Think: Your LLM processes the text and instantly generates an intelligent response.
- Speak: Your TTS (Text-to-Speech) synthesizes the audio and streams it back to the caller seamlessly.
This bidirectional loop runs continuously, supporting ultra-low latency and natural interruptions, entirely over a standard phone call.
Pipeline Diagram
📞 Caller Phone / WebRTC
⚡ TeleCMITransport Stream Bridge
🤖 VoiceAgent Orchestrator
Ready to Build?
The best way to understand Piopiy is to run an agent. Start with the quickstart and then choose your provider stack.
Quickstart
Build and run your first Piopiy voice application.
Explore the Ecosystem
Browse provider options, examples, and recommended stacks.