Harness
Everything around the model that turns it into an agent: tools, system prompt, permissions, hooks.
The harness is the software around the model: the tools it exposes, the system prompt it prepends, how it manages the context window, when it asks permission, and when it stops. Two products running the same model behave differently because their harnesses differ.
This matters for diagnosis. When behaviour differs between two tools, or between yesterday and today, the model is often not the variable. The harness is. A changed system prompt, a different tool set, a new permission default: all change behaviour without touching the model.
It is also where most of your configuration lives. Instruction files, permission settings, and hooks are all instructions to the harness, not to the model.
Related terms
Agent
A model with tools, a system prompt, and a loop that lets it act and observe results.
Tool
A function the model is allowed to call, like reading a file or running a command.
System Prompt
The standing instructions a harness sends on every request.
MCP
A standard way to plug external tool servers into an agent harness.
Guardrails
Checks that stop or constrain an action before it happens.
Want this working inside your team?
It starts with one conversation about your workflow.