What Is an AI Agent Harness?

Jon Noronha says an agent harness is the set of tools an agent uses to do work, check that work, and hand the user something finished, including asking what the user wants. Gamma, a tool for presentations, documents, websites, and social graphics, started in 2020 and took off in 2023. Early on the team made one-off model calls, such as writing one slide or an outline, and wired them by hand into a preset path that could build a 10-slide deck. The change in the last year or two is a higher-level loop: one model does some work, inspects it, uses tools, and keeps going across many calls. Rebuilding Gamma as an agent this year meant building that harness. Gamma is a creative agent, not a coding agent. A coding harness such as Claude Code teaches an agent to read files, run tests, and follow rules such as plan mode. Gamma's tools have to learn an art style, help shape a story, generate images, lay them out, and judge whether they look right.

He disputed the old "GPT wrapper" label, which treats the model as about 90% of the work and the product as a thin back-and-forth. Harnesses show the wrapping is most of the work. The models are similar in their core abilities, and turning them into a product is the harness around them. In his view, OpenAI and Anthropic put a large share of their own effort into harnesses too.