Export MCP Server configuration

[dmantas]

I'm using Langflow Desktop 1.7.1 on MacOS.

I've created a basic flow with an agent component connected to an LLM, an MCP server and Chat input/output.

I'm successfully running the flow on my Desktop.

I want to export the flow so I can run it on my infrastructure, using langflow-runtime k8s Helm chart.

My problem is that in the exported JSON file, the MCP server config is empty.

                "value": {
                  "config": {},
                  "name": "opensearch_mcp_server_dev"
                }
              },

My MCP server is configured like in the screenshot.

I'm trying to define the URL as a Global variable, so in k8s I can populate it using an env var, but the field seems to expect an absolute URL.

How can I move a flow that includes an MCP server to production using the flow export?

Tweaking the json file is not desirable (but I also don't know how to do it, can you let me know? Maybe i can think of something).

Thank you in advance.

[archit-trainee]

also currently https://github.com/langflow-ai/langflow/blob/main/src/lfx/src/lfx/components/models_and_agents/mcp_component.py this depends on mcp config for db which will defeat the purpose of lfx when running/serving single flow when we will be using mcp

[andifilhohub]

hey @dmantas, could you try it in the latest version of langflow?

[dmantas]

Hey @andifilhohub , 1.7.2 is the latest for Desktop, I didn't try it.

However, I installed 1.8.2 Python OSS version and it seems to work fine, i.e. MCP server info is included in the json.

I guess this means that Desktop will also have the fix/enhancement as well. I will come back otherwise.

Thanks a lot, I think it's fine for me!