Serving an agent with the LangGraph CLI dev command

OSS Product Help LangGraph
,
1

I’m using the LangGraph CLI version 0.4.24 to serve an agent with the langgraph dev command. According to the command line help menu (printed below), the dev option is for development. Why is dev just for development purposes? Why should I not use it to serve an agent on a remote server for other people to access via the API?

It looks like Docker is the only other solution for serving the agent. But the dev command doesn’t require Docker so obviously it’s possible to serve an agent without Docker. If I already have Redis and Postgres running on the server where I want to serve the agent, then I see no reason to use Docker for this.

$ uv run langgraph --help

Usage: langgraph [OPTIONS] COMMAND [ARGS]...

Options:
  --version  Show the version and exit.
  --help     Show this message and exit.

Commands:
  build                                📦 Build LangGraph API server Docker image.
  deploy                             [Beta] Build and deploy a LangGraph image to...
  deploy delete                [Beta] Delete a LangSmith Deployment.
  deploy list                       [Beta] List LangSmith Deployments.
  deploy logs                    [Beta] Fetch LangSmith Deployment logs.
  deploy revisions           [Beta] Manage deployment revisions.
  deploy revisions list    [Beta] List revisions for a LangSmith Deployment.
  dev                                  🏃‍♀‍➡ Run LangGraph API server in development...
  dockerfile                       🐳 Generate a Dockerfile for the LangGraph API...
  new                                  🌱 Create a new LangGraph project from a template.
  up                                    🚀 Launch LangGraph API server.
  validate                          ✅ Validate the LangGraph configuration file.
2

It has no horizontal scalability or persistence.

3

I thought the dev server uses in-memory persistence?

4

Anything in prod requires a proper DB, task queue, etc.; dev uses lightweight versions of these that aren’t designed for a distributed system.

5

So using langgraph up is considered a production deployment for a self-hosted server?