Patter

Patter: Give your AI agent a phone number

Patter is an open-source SDK for voice AI that allows developers to connect any AI agent to real phone calls in just four lines of code. Available for Python and TypeScript under an MIT license, the platform provides a fast, opinionated set of primitives for building inbound and outbound voice workflows with sub-500 ms latency.

Key Features

• Minimal Code Requirement: Connect agents to phone calls using only four lines of code, eliminating extensive telephony boilerplate.
• Inbound and Outbound Calling: Handle both incoming calls and outbound dials using the same agent code, complete with native scheduling and voicemail detection.
• Low Latency: Achieves sub-500 ms time-to-first-audio (TTFA) using native 16 kHz streaming and frame-level buffering.
• Flexible Architectures: Supports end-to-end engines like OpenAI Realtime and ElevenLabs ConvAI, or a pluggable pipeline with multiple STT, LLM, and TTS providers.
• Mid-Call Function Calling: Agents can execute tools during calls, such as searching databases, hitting APIs, or transferring to a human.
• Production Guardrails: Includes built-in safeguards like barge-in handling, warm or cold human transfers, and call recording to disk or S3.
• Built-in Dashboard: Monitor live metrics, call history, transcripts, and per-segment costs directly from your machine.
• Local Development Loop: Features a built-in Cloudflare tunnel and terminal test mode to simulate calls from the shell.

Use Cases

• Customer Support: Deploy 24/7 AI phone agents to handle tier-1 support and route calls.

Getting Started

Website: https://www.getpatter.com

Patter provides a fully customizable, self-hosted infrastructure for voice AI without vendor lock-in. By handling telephony, audio streaming, and carrier routing, it enables developers to focus entirely on prompt design, tools, and logic.