Skip to main content

Any Agent Builder Integration

You can use any agent builder with the Agenytics Starter Kit.

Terminology​

  • Agent Service A separate service that exposes a REST API and communicates with the Agenytics service (the current starter kit).

Integration Overview​

If you decide to use an agent builder such as LangChain or Pydantic AI, you need to expose one POST API endpoint on top of your agent service (for example, using FastAPI).

From the Agenytics platform, only the following data is sent to this endpoint:

  • sessionId — can be used as the AI thread, chat identifier
  • chatInput — the user’s chat input

You can refer to the N8N agent implementation as a reference example.

Conversation State Management​

Your Agent Service must maintain its own thread history to support continuous, contextual conversations with the AI.

Agenytics also stores chat history, but only for UI rendering purposes, using different internal formats depending on the UI type. This history is not used for AI context.

In summary:

  • Agent thread history → handled entirely by the Agent Service
  • UI history → stored by Agenytics for display purposes only (Built-in feature)

Response Format​

Your Agent Service must return a response where the output is located at:

data.output

output must be an array of objects, where each object corresponds to one of the supported Agenytics UI types.

Built-in Platform Features​

The following features are handled automatically by Agenytics:

  • Queue processing
  • Frontend polling
  • Parallel chat handling
  • Rate limiting

No additional configuration is required for these capabilities.