Every time I add the MCP node (1st image), all my credentials break (2nd. image). And whenever I redeploy on Easypanel, the MCP node disappears completely (3rd image).
I’ve tested it a few times now, and the issue keeps happening.
Has anyone managed to fix this?
I’m running it in queue mode, using Redis and Postgres.
I’ve seen this kind of thing happen when the deploy setup (like Easypanel) doesn’t keep community nodes between builds. looks like your environment wipes everything and the n8n-nodes-mcp disappears. Then, when it comes back, the credentials break because n8n no longer recognizes the types.
Here are a few things worth checking:
Make sure COMMUNITY_NODES_ENABLED=true is set in the environment before you install the node, and also during redeploy.
Check if the .n8n/nodes folder is getting cleared. If so, you’ll need to persist it with a volume or similar.
If you’re using the official n8n image, try switching to a custom one with the node already baked in (like with n8n-custom)
Also take a look at the container logs after redeploy — if you see something like “unknown credential type”, that confirms the node got wiped.
If you want help setting up a custom build or a quick workaround, let me know.