Dictionary term
Agent Harness
Also called agent scaffold · agent scaffolding
The software layer around a language model that turns it into a working agent: the loop that calls the model, the tools it may use, context and memory management, permissions, sandboxing and the handling of results.
In plain English
Everything wrapped around the AI model that lets it act as an agent: the tools it can use, what it remembers, what it's allowed to do, and the loop that keeps it working.
Analogy
Like the cockpit around a pilot: the controls, instruments and checklists are what let the pilot actually fly the plane.
Why it matters
Two agents on the same model can behave very differently because of their harness, so choosing or building one (Claude Code, the Claude Agent SDK, a managed agent service, or an n8n agent node) is as important a decision as choosing the model.
In practice
Using Anthropic's Claude Managed Agents as a hosted harness for a long-running onboarding agent, instead of writing your own loop, tool execution and session storage.
Related terms
Sources
Browse all terms →Last verified