---
name: open-flow
description: Create, inspect, edit, check, run, and publish persistent Open Flow workflows through an existing Server, or assess and convert n8n workflows to Open Flow. Use for workflow operations, not changes to the Open Flow implementation.
---

# Open Flow

Use `<!-- agentic:var flowCommand -->` to operate the selected deployment.
Consult the references relevant to the task:

- [Flow capabilities and examples](references/flow-authoring.md): inspect, edit,
  connect, validate, run and publish; concurrency and Run diagnostics.
- [n8n conversion](references/flow-n8n-conversion.md): represent collections,
  branches, provider operations and state while preserving source behavior.

<!-- agentic:if localDevelopment -->

Run commands from this repository's root. The development CLI connects to the
existing local Server by default, using `OPEN_FLOW_PORT` (default `3001`) and
`OPEN_FLOW_TOKEN`, or the token generated by `bun run dev`. It reads the root
`.env`. If the local Server has not been started, report that prerequisite.
Do not start or reconfigure a Server merely to perform a Flow operation.

To select another Server, the user must provide both `OPEN_FLOW_URL` and
`OPEN_FLOW_TOKEN` in the environment. Never read, print, or copy credential files
into prompts or command arguments. Connector and LLM configuration stay on the
Server; do not duplicate them in the CLI.
<!-- agentic:endif -->

Keep the requested boundary: editing a Draft does not authorize a Run,
publication, or enabling automatic execution. Report the resulting Flow and
Revision, verification result, and any unresolved configuration.