Traditional agent development splits behavior across prompt templates, tool schemas, callbacks, and workflow graphs, forcing developers to learn new abstractions. NOOA instead represents an agent as a single Python class: its methods are the actions the LLM can take, fields are state, docstrings become prompts, and type annotations serve as contracts. Methods whose body is '...' are completed at runtime by an LLM‐driven loop; methods with normal bodies remain standard deterministic Python, giving developers and agents the same interface.
The framework is model‐agnostic and exposes agent‐specific capabilities (context, events, long‐term memory, validated LLM loops) through simple Pythonic APIs. NOOA combines six model‐facing ideas on one surface: typed input/output, pass‑by‑reference over live objects, , programmable loop engineering, explicit object state, and model‐callable harness APIs. The authors survey fourteen other frameworks and find the community already converging on several of these ideas.
Evaluation on SWE‑bench Verified, Terminal‑Bench 2.0, and ARC‑AGI‑3 shows that current models use the interface effectively. On ARC‑AGI‑3, the framework compresses a multi‑agent world‑model system into a single agent with a one‑page skill while advancing the benchmark’s score–cost Pareto frontier. The approach is interesting because it eliminates the learning curve for humans and ensures immediate agent readiness by leveraging familiar Python abstractions.