Skip to main content

Artemis Documentation

Artemis is a code optimisation platform. You give it a codebase, an Objective, a representative benchmark and the checks a change must pass. Artemis generates candidate changes, runs experiments on hardware you control, compares the results with the Baseline, and returns the improvements as reviewable code changes with their evidence. You can use it in the browser, from the Artemis CLI, or from a coding agent such as Claude Code.

Inside a Project​

The Project sidebar follows the work: Overview, AI Agents, then the workflows Maintain, Discover and Plan, then the code in Branches and Files, and finally Project Settings.

Maintain​

Define Rules, scan your code for Issues, triage the findings, and fix them with a coding agent, in a Discovery Run, or with your own local agent. Rules can be written in Markdown, drafted with the Maintain agent, copied from another Project or taken from the Rule Library.

Discover​

Run a Discovery: Artemis explores many candidate changes toward an Objective with several models, measures each Version against the Baseline on your Runner, and keeps the evidence for every Experiment. Without a Runner, a panel of judge models scores the Versions instead, for work such as improving a written Plan.

Plan​

Break a piece of work into Tasks before any code is written, then build the Plan on a new Branch.

Branches​

Review, verify and publish proposed code changes. Every coding agent works on its own Branch, and nothing reaches your Git provider until you publish it.

AI Agents​

Start a chat, a Plan, a Maintain agent or a Discovery Run, and follow every agent working on the Project from one page.

Get started​

Platform and settings​

Deployment and architecture​

Need help?​

Last updated on by Mike Basios