NVIDIA-labs OO Agents: Native Python Object-Oriented Agents

2026-07-22

AI Summary

This work addresses the lack of a unified programming model in traditional agent development, which is fragmented across prompts, tools, and workflows. The authors propose an object-oriented agent framework that models agents as native Python objects, where methods represent actions, fields encode state, docstrings serve as prompts, and type annotations define contracts, enabling seamless hybrid execution of LLM-driven and deterministic code. For the first time, this framework integrates six key interaction features within a single system: typed I/O, live object reference passing, code-as-action, programmable loops, explicit state management, and context- and event-aware APIs callable by language models. Designed with a purely Pythonic philosophy, it unifies developer and agent interfaces. Empirical validation on benchmarks including SWE-bench Verified, Terminal-Bench 2.0, and ARC-AGI-3 demonstrates its effectiveness in significantly enhancing large models’ task execution and reasoning capabilities.

Abstract

Traditional agent development is split across prompt templates, tool schemas, callback code, and workflow graphs. We present NVIDIA Object-Oriented Agents (NOOA), a model-agnostic Python framework for building reliable AI agents. NOOA takes a simpler approach: an agent is a Python object. Its methods are the actions the model can take, fields are its state, docstrings are its prompts, and its type annotations are contracts. A method whose code body consists of "..." is completed at runtime by an LLM-driven agent loop, while methods with normal bodies remain standard deterministic Python. This gives developers and agents the same interface, so agent behavior can be tested, traced, refactored, and improved just like other software. This paper makes three contributions. (1) We present the agent-as-a-Python-object programming model and the design principles behind it. Where Python has existing abstractions, we adopt them directly. Agent-specific capabilities--context, events, state rendering, long-term memory, and validated LLM loops--are exposed through simple Pythonic APIs, so both developers and agents share one familiar programming model. (2) We identify six model-facing ideas that NOOA is, to our knowledge, the first to combine on a single surface: typed input/output, pass-by-reference over live objects, code as action, programmable loop engineering, explicit object state, and model-callable harness APIs for context and events. We find the community already converging on several of these ideas--often as experimental or partial features--and present the comparison to encourage further adoption. (3) We demonstrate that current models use this interface effectively, both in targeted capability tests and on agentic and reasoning benchmarks such as SWE-bench Verified and Terminal-Bench 2.0 and ARC-AGI-3.

Problem

• AI agents
• programming model
• software engineering
• object-oriented
• LLM integration

Innovation

• Object-Oriented Agents
• Python-native AI agents
• LLM-driven runtime completion
• Typed I/O and stateful objects
• Programmable agent loops