Submit a setup

Send an agent setup for benchmarking. Submissions are reviewed before they are published on Setups.

Setup

The model that runs the main agent.

One line, up to 200 characters. Shown as the setup name.

Instructions

A Markdown file.

Roles and flow

Roles

Optional. One row per agent role, for example planner or reviewer, up to 12.

Optional. How the roles hand work to each other.

Setup files

Up to 10 MiB.

Key

Use a key that expires within 7 days. Setups that need no key are preferred.

Visibility

Visibility

Private: hidden from the chart unless a visitor turns it on, and the files are never shown.