Docker

docker run -d --name rookery -p 8080:8080 \
  -v rookery-data:/data ghcr.io/rookery-ai/rookery:latest

Open http://localhost:8080.

Create the owner account

docker exec -it rookery rookery owner bootstrap -u yourname -p 'a-long-password'

The one limitation

The image ships no command-line coder tool. A workspace running in the container must connect to a model provider directly rather than driving a local coder. See Choosing a model.

This is deliberate — bundling a coder tool would multiply the image size — and it is enforced rather than merely undocumented: the container refuses to be configured for a coder it does not contain, and tells you so.

Your data

Everything lives in the volume mounted at /data: the database, every workspace’s knowledge base, and the encrypted credentials.

docker volume inspect rookery-data

Back it up. The volume is the installation.

Configuration

Environment variables work the same as a native install:

docker run -d --name rookery -p 8080:8080 \
  -v rookery-data:/data \
  -e ROOKERY_PUBLIC_URL=https://rookery.example.com \
  ghcr.io/rookery-ai/rookery:latest

Health

docker exec rookery rookery healthcheck

Reports version, protection status, and which optional tools are present. The image includes the tools Rookery uses for reading PDFs, spreadsheets and scanned documents, so a healthy container reports no missing-tool warnings.