# Harness Engineering

> The practice of designing, testing, and simplifying the runtime around an agent so it can work reliably across tools, sessions, and tasks.

- Category: Agents & Autonomy
- Stage: Rising
- Also called: Agent harness engineering, Harness design
- Canonical: https://www.joelcsaji.com/ai-dictionary/harness-engineering

## Plain English

It is systems engineering for the part around the model: context, tools, permissions, state, loops, and evaluation.

## Example

A team improves a coding agent by adding sprint contracts, a QA evaluator, Playwright checks, and then removing pieces once a newer model no longer needs them.

## Why it matters

Recent agent gains often come from better runtime structure, not only from larger models or better prompts.

## Related

- [Agent Harness](https://www.joelcsaji.com/ai-dictionary/agent-harness)
- [Context Engineering](https://www.joelcsaji.com/ai-dictionary/context-engineering)
- [Agent Loop](https://www.joelcsaji.com/ai-dictionary/agent-loop)

## Sources

- [Anthropic: Harness Design](https://www.anthropic.com/engineering/harness-design-long-running-apps)
- [arXiv: ALARA for Agent Harness Engineering](https://arxiv.org/abs/2603.20380)
- [OpenAI: Agents SDK Evolution](https://openai.com/index/the-next-evolution-of-the-agents-sdk/)