Skip to main content
The Atoms SDK is built around a graph-based architecture where information flows between processing units. This design makes it easy to build everything from simple chatbots to complex multi-agent systems.

​
How It Works

Every voice conversation in Atoms follows this pattern:
  1. Audio comes in — The user speaks into their phone or browser
  2. Events flow through nodes — Speech is transcribed, processed, and responses generated
  3. Audio goes out — The agent’s response is synthesized and played back
The system manages all the complexity of real-time streaming, interruptions, and state management. You just focus on the logic.

​
The Four Building Blocks

Nodes

Processing units that handle events. The brain of your agent logic.

Events

Messages flowing through the system. Audio, text, and control signals.

Graphs

Connect nodes into pipelines. Build complex multi-agent flows.

Sessions

Manage conversation state and lifecycle. One session per call.