ax-agent

You can build a multi-agent system by using a fluent API to define agents with structured inputs and outputs, then composing them into hierarchies with shared fields and integrated external tools for orchestrated execution.

Recursive language model (RLM) processing handles arbitrarily long documents by breaking them down into manageable code execution and LLM queries, enabling efficient long context processing without losing information.

You integrate external tools into an AI agent framework by defining agents with a fluent API and incorporating external tools directly, allowing agents to perform actions like web searches and return structured outputs.

Yes, you can run sandboxed code execution environments within agents, which is specifically used during recursive language model processing to safely execute code when breaking down long documents.

Yes, the framework supports shared fields across multiple agents, enabling global agents within sophisticated hierarchies to access common data and coordinate complex multi-agent workflows effectively.

The best way to process arbitrarily long documents is using RLM mode, which breaks down long texts into manageable code execution and LLM queries to handle extensive context efficiently.