Try Artemis

The fastest way to judge Artemis is to use it. The setup below includes the integrated code lifecycle, so it covers programming exercises: Artemis hosts the git repositories itself and runs the tests in Docker containers on a build agent. That is the part of Artemis worth looking at first.

Run it locally with Docker

You need Docker with the Compose plugin, and about 6 GB of memory available to it (in Docker Desktop: Settings → Resources). Published images cover both amd64 and arm64, so Apple Silicon pulls a prebuilt image rather than building one.

git clone https://github.com/ls1intum/Artemis.git
cd Artemis
# Linux with a native Docker daemon only: the build agent needs your host's docker group to reach the
# Docker socket. Compose resolves this when it creates the container, so it has to be set beforehand.
export ARTEMIS_DOCKER_GROUP_ID=$(getent group docker | cut -d: -f3)
docker compose --env-file .env -f docker/artemis-dev-local-vc-local-ci-postgres.yml pull
docker compose --env-file .env -f docker/artemis-dev-local-vc-local-ci-postgres.yml up

Then open http://localhost:8080 and sign in as artemis_admin / artemis_admin.

Docker Desktop on macOS and Windows needs no docker group, so the export line is harmless there and can be skipped.

⚠️

Warning:

Run Compose from the repository root and keep --env-file .env. The Compose files read image versions from the .env file in the repository root, and Compose looks for that file next to the first -f file, which is docker/, not the root. Without it the images resolve to blanks and the project either fails to start or fails to pull.

To stop, press Ctrl+C. To remove the containers and the database volume as well:

docker compose --env-file .env -f docker/artemis-dev-local-vc-local-ci-postgres.yml down -v

What you can do with it

The first start creates the database schema, so give it a moment. Once Artemis reports itself started, you have a complete instance:

• Programming exercises. Create one and Artemis provisions the template, solution, and test repositories on its own git server. Students participate, push commits, and the build agent runs the tests in a container and reports feedback. The online code editor, manual code review, and submission policies are all part of this setup.
• Every other exercise type. Quiz, text, modeling, and file upload, with manual assessment against structured grading criteria.
• Exams. Including exercise variants per student and instructor test runs.
• Courses, lectures, competencies, learning paths, communication channels, and FAQs.

The setup starts one build agent with four concurrent build slots, which is enough to feel how the feedback loop behaves. Production deployments scale this out horizontally.

ℹ️

Information:

The build agent pulls a language-specific Docker image the first time an exercise of that language is built, so the very first build takes noticeably longer than the ones after it, and it needs working network access to the image registry. If a build does not appear, look at Server Administration → Build Agents and the build queue: the agent has to be listed and reachable, and it needs access to the Docker socket that the Compose file mounts for it.

What it deliberately leaves out

• AI features. Iris, Athena, and Hyperion are optional integrations and need a configured model provider. See Artemis Intelligence.
• External authentication. The trial uses internal accounts. LDAP, SAML2, and OIDC are configured per deployment.
• Mail. Notification emails need an SMTP server; Mailpit is the local option.

❗

Danger:

This is a trial setup, not a deployment. It runs the development profile with the credentials published in this repository, and it gives the container access to your Docker socket so the build agent can start containers. Never expose it to a network you do not control. For a real deployment, follow the administrator guide; the production profile refuses to start until you replace those credentials with your own.

Simpler variant, without programming exercises

If you do not need programming exercises, or you would rather not mount the Docker socket, this one is smaller and needs no group configuration:

docker compose --env-file .env -f docker/artemis-dev-postgres.yml up

Ask for a hosted trial

If you would rather not run anything, or your institution is too small to operate its own instance, the Artemis team at TUM can host one for you. This is intended for schools and for institutions evaluating Artemis before committing to their own infrastructure.

Write to artemis@xcit.tum.de.

Deploy it properly

When the trial has answered your questions:

• Administrator guide for production setup, security, scaling, and operations.
• Integrated code lifecycle setup for running programming exercises in production, including separate build agents.
• Artemis Ansible Collection automates a production deployment and is the recommended path.
• Docker Compose setups documents every Compose file in the repository.
• Trust and transparency answers the security, privacy, accessibility, and AI questions that come up in an institutional review.

Develop against it

If you want to change Artemis rather than only run it, start with the developer setup guide, which sets up the server and client for local development with hot reload.