NodeFox
Deployment

Run it in the browser,
or on your own machines.

Build and run in the browser with nothing to install. Export compatible workflows to supported Linux or macOS runners: on a schedule, when files land or from a webhook.

Execution modes

One workflow, two places to run it.

Confirm that the workflow’s steps, connections and operator inputs are supported headlessly. Headless execution is included with Pro and Enterprise.

Browser
Headless Pro & Enterprise
Runs on
The browser session of the person running it
Your Linux or macOS machine, as a background service
Starts from
A person, an app form or a batch
Command line, schedule, watched folder or webhook
People decide
In the NodeFox app, when a run awaits input
From the operator’s terminal or an attached agent
Suited to
Building, reviewing and attended work
Unattended volume on a schedule or trigger
Headless

How a headless deployment works.

The runner, nf, is a background service on your machine under launchd or systemd. It starts at login and restarts after a crash.

  1. 01

    Send it from the browser

    Export the App or Automation as a file, or send it straight to a paired runner.

  2. 02

    Install the runner

    One command with an install code downloads nf and signs the machine in.

  3. 03

    Choose a trigger

    Run it from the command line, on a schedule, from a watched folder or from a webhook.

  4. 04

    Read the record

    Every run leaves a record on the machine: what ran, what it produced and how it ended.

terminal
# install and sign in with your install code
curl -fsSL https://nodefox.ai/nf/install.sh | sh -s -- --code NF-…

# deploy what you exported, run it as a service
nf deploy order-check.nfx
nf service install

# weekdays at 07:00, Chicago time
nf dispatch order-check --schedule "0 7 * * 1-5" --tz America/Chicago

# what ran and how it ended
nf runs
Triggers

Start runs the way your work arrives.

Each deployed workflow has one trigger, with its own policy for errors and for questions nobody is there to answer.

Command line

Run on demand, from your own scripts, or from an agent through the runner’s MCP server.

Schedule

Cron with a time zone; a firing missed while the machine was off is skipped, not replayed.

Watched folder

Each file that lands in the folder starts a run with that file as its input.

Webhook

A bearer-token endpoint on the machine’s loopback; put your reverse proxy in front to expose it.

What your team provides

A machine, access and keys.

A host machine

Linux x64 or arm64 with glibc 2.29 or newer, or an Apple Silicon Mac.

A reverse proxy with TLS

Only if a webhook must be reachable from outside the machine.

Access to your systems

API credentials or file locations for what the workflow reads and writes.

Model provider keys

Your own keys for the providers the workflow calls; they stay on the runner.
Not available today
  • Windows and Intel Mac runners
  • Offline or air-gapped runs
  • MCP tools and OAuth connections inside headless workflows
  • Remote or mobile approvals for headless runs
  • A runtime or control plane hosted for you

Does a headless run need the internet?

Yes. Each run fetches a short-lived run key from NodeFox, and model requests go to your provider.

What happens if the machine restarts or the runner crashes?

The service starts again on its own. A run cut off part-way is marked interrupted and never re-run automatically, so you decide what happens next.

Who answers a question a headless run asks?

Its operator: from the terminal, from an MCP client such as an agent over SSH, or from a browser paired on the same machine.

What happens when a step fails?

You choose per workflow: stop, ask the operator, or retry up to ten times.

Plan where your workflow will run.

Bring the workflow and the machine you have in mind. We’ll cover triggers, access and what each run should record.