Editor Quickstart
Run the visual editor from Docker and execute a flow from the UI.
The standalone editor is a published Docker image that bundles the visual flow
editor and the octo runtime. You get a browser-based canvas for designing
flows and a Run button that executes them in the container — no build, no
database, no account.
Start the container
docker run -p 3000:3000 -v "$PWD:/work" juancavallotti/octoThe volume mount is the important part: the editor reads and writes flow YAML
as plain *.yaml files under /work, so -v "$PWD:/work" makes your current
directory the flow store. Your flows live on your disk, in your directory —
stop the container and the files are still there, ready for octo run or a
git commit. Mount an empty directory to start fresh, or the cloned repo to
browse its samples/.
Open the editor
Go to http://localhost:3000. The file menu in the header lists every YAML file in the mounted directory. Open one, or create a new flow — the canvas shows the flow's sources and blocks, the palette on the left holds the available blocks, and selecting a block edits its settings.
Save and run
Save writes the flow back as a .yaml file in the mounted directory. Hit
Run and the bundled octo binary starts the flow inside the container with
hot reload — edit, save, and the running flow picks up the change. Logs from
the run stream into the editor so you can watch messages move through the flow.
If your flow serves HTTP, the editor proxies it for you: a test URL appears in
the console, so you don't need to publish extra ports. (Publishing a port, for
example -p 8080:8080, is only needed when you run octo run yourself inside
the container.)
If a flow declares environment variables, the editor's Dev .env panel edits
their values. They are stored in a shared .env.dev file next to your flows,
and runs read them from there.
The image sets OCTO_FS_DIR=/work, which is why the mount target matters; you
only need other environment variables when you go beyond the defaults. It also
exposes an MCP server at http://localhost:3000/mcp so an AI assistant can
draft, validate, and run flows against the same files — see
MCP authoring.
For the full tour of the editor — how it works, running flows, and configuration — see the Standalone Editor section.