Flowdrome is a workflow platform that compiles every flow into a standalone executable and runs it on your own fleet. Build visually, test live against real data, then ship a binary — to a bare server, a container, or a VM that boots straight into being a host.
A -drome is a place where things run — velodrome, aerodrome, Flowdrome.
The Studio editor: typed ports, live validation, single-step debugging and the whole run envelope on every wire.
Every workflow compiles to a single standalone executable (Bun-compiled, dependency-free). Your automation is a program you can copy anywhere — not rows in someone else's queue.
The Nucleus manages workflows, credentials, users and deployments; hosts execute. Enroll a host, push a build, watch every deployment from one place — the Nucleus itself never runs a workflow.
Triggers, flow control, list tools, SQL/S3/Redis/Mongo/FTP/IMAP wire clients, Telegram,
Slack, Discord, Google Sheets/Docs/Drive/Gmail, Notion, HubSpot, Outlook, AI nodes — plus
universal {{ }} expressions in every string field.
The Approval node pauses a run until a person decides. Pending gates from every host land in one fleet-wide inbox in the Nucleus — approve, reject, or let the configured timeout policy decide.
Credentials are encrypted at rest in the Nucleus, with OAuth2 connect flows for Google, Notion, Microsoft and HubSpot. Deployed apps lease secret values at run start over a scoped token — nothing is ever baked into an artifact.
Every run records the whole envelope on every wire — inspect any node's exact input and output, per port. Hosts keep a run ledger and resource telemetry; retry a failed run from the node that broke.
Most platforms execute your flows inside their runtime, forever. Flowdrome's compiler emits a self-contained executable per workflow — start it with an init system, ship it in a container, run it air-gapped. The editor and the artifact share one execution model.
The live appliance ISO is the operating system: boot it and the machine is a
Flowdrome host in seconds — no installer. It adopts a data disk labelled
NINER_DATA (or provably blank) for persistence and dials home to your
Nucleus.
One Nucleus manages many hosts; hosts run workflows and keep running if the Nucleus goes away. Management traffic and workflow traffic never share a process — or a failure domain.
Deployed apps hold credential references, never values. At run start they fetch the real values from the Nucleus over a short-lived, deployment-scoped lease and keep them only in memory.
Drag nodes onto the canvas, wire typed ports, template any field with
{{ expressions }}. Validation lights up problems before you ever run.
Run against real services from the editor. Inspect the full envelope on every wire, single-step the debugger, edit data mid-run, retry from any node.
One click compiles the workflow to an executable and pushes it to a host in your fleet. Real triggers take over; the Nucleus watches every deployment.
From git clone to a deployed workflow in one sitting.