Agents & tools
Agent Harness
The scaffolding around an LLM — tools, memory, loops, and orchestration — that turns a model into an agent.
01 ——
In plain English
An agent harness is the software layer that wraps a language model and gives it the abilities a true agent needs: a tool-calling loop, working memory, error handling, and a way to stop when the goal is met. The harness is what separates a one-shot chatbot from something that can actually do work.
What a typical harness provides:
- Tool-use loop — call → run → feed result back → call again, until done
- Memory — short-term context and longer-term storage of facts or files
- Planning — break a goal into steps, often with a separate planner model
- Guardrails — limits on tool calls, spending, or unsafe actions
- Observability — logs, traces, and replay for debugging
Examples: Claude Code, Cursor, OpenAI's Agent SDK, LangGraph, CrewAI, and Anthropic's Claude Agent SDK are all harnesses. The underlying model (Claude, GPT, Gemini) is interchangeable; the harness is what creates the user-facing product.
02 ——
Related terms
AI Agent
Software that uses an LLM to plan and act — picking tools, taking actions, and adapting based on results to complete a user’s goal.
Agent Orchestration
The coordination layer that decides which agent or tool runs next, manages state across steps, and handles failures in multi-step AI workflows.
Tool Use
The ability of an AI model to call external tools — like a calculator, search engine, or API — to help answer a question.
Function Calling
A feature that lets an AI model trigger specific functions or APIs in your app instead of just returning text.
Coding Agent
An AI agent specialised for writing, editing, and debugging code — usually with the ability to read your repo, run tests, and open pull requests.
Agentic Workflow
A multi-step task where an AI agent autonomously decides the steps, uses tools as needed, and works toward a goal with minimal human steering.
Back to glossaryLast reviewed May 2026