Definition
An AI Agent is an orchestrated computing system pairing a foundational reasoning model (LLM) with an external runtime harness that provides tools, context, long-term memory, and environment execution privileges.
Purpose
Enables probabilistic reasoning models to execute deterministic operations, read live software environments, and alter real-world system states.
Key Concepts
How It Works
Formulated systematically as AI Agent ≈ Model + Harness. The model acts as the reasoning brain, while the harness provides tools, data, and feedback loops.
Related Terms
Sources
- Birgitta Böckeler, AI Agent Harness, Thoughtworks
- Harrison Chase, LangChain Agent Architecture