Prepatu
Define voice agents in YAML. The state machine runs the show. The LLM does the talking.
The Problem
Generative models excel at natural language comprehension but inherently struggle with deterministic control flow. When deploying multi-step voice agents, relying on probabilistic models for logic results in skipped states, hallucinated confirmations, and critical compliance failures.
Every enterprise team building voice agents hits the same wall: probabilistic models cannot be trusted with rigid flow control.
The Solution
Prepatu separates concerns:
YAML State Machine
Defines the flow: states, transitions, tools, artifacts. Declarative and version-controlled.
LLM as Guest
Operates inside each state for natural language understanding, entity extraction, and tool calls.
Engine as Host
Enforces the rules. The LLM cannot skip states, call foreign tools, or invent transitions.
Flexible Deployment Architecture
☁️ Managed Cloud Platform
Zero-infrastructure deployment with enterprise SLAs and global edge delivery.
📦 Self-Hosted Enterprise
Complete data sovereignty, custom VPC integration, and full infrastructure control.
🧩 Native SDK Integration
Seamless implementation into existing frontend architectures with robust type safety.