# What is agent harness?

> An agent harness is everything in an AI agent except the model: the tools, context, prompts, memory, hooks, guardrails, and feedback loops that turn a model into a working agent.

- HTML: https://openlit.io/glossary/agent-harness
- Markdown: https://openlit.io/glossary/agent-harness.md

An AI agent is a model plus a harness. The model generates tokens; the harness decides which tools to expose, how context is assembled, what permissions apply, and how failures are handled.

Common harnesses include coding agents such as Claude Code, Codex, Cursor, and Windsurf, plus frameworks such as LangGraph, CrewAI, OpenAI Agents SDK, and deepagents. Each is a different way to wrap a model with tools and control flow.

OpenLIT does not replace your harness. It instruments any harness through OpenTelemetry so you can observe, evaluate, and improve it: traces for LLM and tool calls, evals on real trajectories, guardrails at runtime, and prompt or rule changes without redeploying.

## FAQ

### What is an agent harness?

An agent harness is everything in an AI agent except the model: tools, context, prompts, memory, hooks, guardrails, and feedback loops.

### Is LangGraph an agent harness?

LangGraph is an agent framework you use to build a harness. The running system around the model—tools, state, and control flow—is the harness.

### How does OpenLIT relate to an agent harness?

OpenLIT observes, evaluates, and improves any harness via OpenTelemetry. It is not a runtime harness itself.

## Related

- [Harness engineering](https://openlit.io/glossary/harness-engineering.md)
- [Agent observability](https://openlit.io/glossary/agent-observability.md)
- [Agent evals](https://openlit.io/glossary/agent-evals.md)
- [Guardrails](https://openlit.io/glossary/guardrails.md)
- Pillar: https://openlit.io/agent-harness-engineering.md