</>MCP Agents Market
Agent

labs-OO-Agents

by NVIDIA-NeMo1.8kPythonUpdated 2026-08-21

NVIDIA Object Oriented Agents: the Pythonic way to build AI Agents.

NVIDIA Object Oriented Agents (NOOA) is a model-agnostic Python framework that enables developers to build AI agents using familiar object-oriented programming patterns. It allows agents to be expressed as Python classes where methods with ellipsis bodies become LLM-driven generation methods, while regular methods remain deterministic Python code. The framework supports typed interfaces, live object state, and code-as-action execution in a Jupyter-style REPL, making AI agent development integrate naturally with standard Python workflows including testing, version control, and refactoring.

Key Features

Object-oriented agent design where Python classes define state, capabilities, prompts, and typed contracts in one interface
Generation methods with ellipsis bodies that are implemented at runtime by LLM-driven strategies
Code-as-action execution where models write Python code in a REPL environment with access to self, imports, and helper methods
Built-in tracing for every LLM call, code execution, and method invocation with parent-child span tracking
Model-agnostic design supporting LiteLLM-compatible providers including Anthropic, OpenAI, Ollama, and vLLM
Typed input/output with automatic retry, live-object arguments passed by reference, and model-callable context APIs
Optional CLI, memory subsystem, Agent Client Protocol integration for editors like Zed, and benchmark harness
AST validation and module deny-lists for defense-in-depth safety guardrails on generated code

Use Cases

  • 01Building customer support agents that analyze feedback, triage tickets, and check business rules programmatically
  • 02Developing coding assistants that execute Python code and interact with live objects in development workflows
  • 03Creating structured data extraction agents with typed output contracts and automatic validation
  • 04Implementing multi-agent systems where subagents are composed as Python objects with clear interfaces
  • 05Prototyping research agents with traceable execution for benchmarks like SWE-bench Verified and Terminal-Bench
  • 06Integrating AI agents into existing Python applications using standard software engineering practices

Related Agents

View more

labs-OO-Agents — FAQ

What is NVIDIA Object Oriented Agents?+

NVIDIA Object Oriented Agents (NOOA) is a Python framework for building AI agents using object-oriented design where agents are Python classes, methods define capabilities, and LLM-driven logic is expressed through type-annotated methods with ellipsis bodies.

How do I install NVIDIA Object Oriented Agents?+

Install using uv with 'uv add nooa' or pip with 'pip install nooa'. Optional packages like nooa-cli, nooa-memory, and nooa-acp can be installed separately or as extras like 'uv add nooa[cli,memory]'.

What LLM providers does NOOA support?+

NOOA is model-agnostic and works with any LiteLLM-supported provider including Anthropic Claude, OpenAI GPT models, local Ollama models, and self-hosted vLLM servers.

Do I need API keys to use NOOA?+

API keys are required for hosted providers like Anthropic (ANTHROPIC_API_KEY) and OpenAI (OPENAI_API_KEY). Local models via Ollama or vLLM do not require API keys.

Is NVIDIA Object Oriented Agents free?+

Yes, NOOA is open-source software released under the Apache 2.0 license and free to use. However, LLM provider costs apply when using commercial APIs.

Is it safe to run agents that execute LLM-generated code?+

NOOA includes AST validation and deny-lists but these are defense-in-depth measures, not containment boundaries. Always run agents in sandboxed environments like containers, VMs, or NVIDIA OpenShell to prevent unwanted file access or data exfiltration.

How do I install labs-OO-Agents?+

Open the source repository on GitHub and follow its README. labs-OO-Agents is a agent — MCP Agents Market links you directly to the official repo.

Is labs-OO-Agents free?+

labs-OO-Agents is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches