HAIP serves as a layer that wraps all communication between your Agents and the User client. It manages connections automatically, letting you focus on development while ensuring speed and reliability. HAIP aims to be flexible and allow you to develop your solutions easily.
  • Plug-in your existing identity providers.
  • Service agnostic. Expose multiple agents/LLM pipelines.
  • Create your own schemas.
  • Run on your own infrastructure.
  • Credit management.

​
Architecture Overview

There are two parts to HAIP. The frontend code which will get pulled as a library and used in your frontends (eg. app, web and voice). This will then communicate using the HAIP protocol to the HAIP server running on your infrastructure.

​
Why to use HAIP?

This protocol can be thought of a standard way to call LLM-like workflows. And the SDK is like React Query for LLM use cases. It’s meant to allow you to set up connections to your LLM in an easy to manage way. While having all the aspects that you would otherwise have to build like chat histories, credit management and flow control. An example of a client: using our drop in React libraries. 
This would be an example of the React Library

​

Read the protocol overview

Explore the complete HAIP API protocol specification.

​
Best Practices

Error Handling

Implement comprehensive error handling for network issues and protocol violations.

Flow Control

Use flow control to prevent overwhelming your system with messages.

Authentication

Always validate JWT tokens and implement proper session management.

Monitoring

Monitor connection health and message throughput in production.

​
Development Tools

Development Status: Some development tools are still in development. The SDK itself is fully functional and can be used for development and testing.

​
HAIP CLI