Agent Harness
The software around a language model that lets it act: it lists the tools the model may use, reads the tool calls in the model's reply, runs them, and feeds the results back into the next turn of the Agentic Loop. Most harnesses also manage instructions, Agent Memory, permissions, context and stop conditions. A model plus a harness is an agent; Claude Code, for example, is an app built on the Claude Agent SDK harness. The same model can be unreliable in one harness and dependable in another.
Written by Max Zeshut
Founder at Agentmelt · Last updated Oct 26, 2026
Example
A support team's agent drafts refunds. The harness gives it three tools (look up the order, check the refund policy, issue a refund), runs each call the model asks for, and stops any refund above $200 for a person to approve. Swapping in a newer model changes nothing about those rules, because they live in the harness, not the model.
Frequently asked questions
- Is a harness the same as an agent framework?
- No. A framework (LangGraph, the OpenAI Agents SDK) is a library for building a harness. The harness is the running result: a specific set of tools, rules, memory and checks wrapped around a model.