@TLeM4 said: Issue: ImageMagick v6 provided and v7 required. That was it indeed.

Following up on this thread with the state on the current package (1.18.3 / Penpot 2.16.2). The flags @james added are working, and /mcp now resolves. However, the in-app MCP plugin still can't load, which blocks File → MCP Server → Connect. This is exactly the missing piece @Operator described (the frontend plugins/ static assets aren't served). Current state on 1.18.3: [v] enable-mcp + enable-access-tokens are in /app/pkg/start.sh and served in /js/config.js (penpotFlags). [v] /mcp now returns 200 (progress vs the earlier 404s in this thread). In a design file, File → MCP Server → Connect fails. Browser console: Plugin Error: Failed to load plugin code GET https://penpot.inno3.eu/plugins/mcp/plugin.js 404 (Not Found) Root cause (confirmed): the frontend loads the bundled MCP plugin from ${PENPOT_PUBLIC_URI}/plugins/mcp/…, but the nginx root /app/code/frontend has no plugins/ directory -- the files don't exist anywhere in the container. The upstream image already ships them: $ docker export $(docker create penpotapp/frontend:2.16.2) | tar -tf - | grep plugins/mcp var/www/app/plugins/mcp/ var/www/app/plugins/mcp/assets/index-JunN9mS-.css var/www/app/plugins/mcp/icon.jpg var/www/app/plugins/mcp/index.html var/www/app/plugins/mcp/index.js var/www/app/plugins/mcp/manifest.json var/www/app/plugins/mcp/plugin.js For reference https://design.penpot.app/plugins/mcp/plugin.js returns 200; ours returns 404. Related: our config.js exposes none of the penpotPlugins* vars (penpotPluginsWhitelist, penpotPluginsListUri, …), so the Plugins manager panel is unavailable too for same root cause. As @Operator noted, /app/code is read-only at runtime, so this needs to be handled at image-build time. It looks like copying the upstream frontend plugins/ directory into the served root (and exposing the PENPOT_PLUGINS_* config in config.js) would light up both the native MCP plugin and the Plugins manager, without needing to bundle the separate penpot-mcp Node service. Happy to test a patched package on our instance.

@james give me a couple of days and i shall revert.

Much appreciated!

@james I haven't tried to make the modifications suggested since then, but i have good reasons to believe it'd work.

After a server migration I expirienced a somehow similar behaviour. After the cloudron restored from the backup penpot OIDC didn't work. penpot package version: 1.7.0 Frontend said: "Authentifizierungsanbieter ist nicht konfiguriert." On the backend: "E app.http.errors - hint="restriction error", data={:type :restriction, :code :provider-not-configured, :provider :oidc, :hint "provider not configured“}" I did the following diagnostics: I checked OIDC login on other apps (e.g. gitlab) -> worked as expected. I checked curl -v https://my.mydomain.de/.well-known/openid-configuration -> it answered correctly cloudron-support —troubleshoot -> said all checks [OK] I freshly installed another instance on the same machine but different domain. As a regression to 1.1 was no Option -> same error I double checked DNS entries. -> same error I renewed the password of the accounts setup in penpot -> same error Now after about 24h after the restore I tried again a fresh install -> it worked Returned to the previous instance –> It all of a sudden worked. Solution: Time is a healer… This is realy very astonishing and I feel lucky, but it is not very reliable and now brings up questions how to secure the designwork of me and others when working with penpot. Unfortunatly I have no time left to further investigate, but it seems that OIDC & penpot have issue after restoring of cloudron...

Ok! Thank you! Cloudron in any case amazing product

press play & get link -> https://penpot.example.com/#/view/…

It is currently not on our list, given the penpot installation count we see and that being quite the edge case on top which needs extra maintenance on our side. We will keep it in mind for when we have free resources to tackle that.

Ah I didn't realize this is the actual domain you had installed Penpot, so never mind my comment about external resources not loading. What makes you think your issue is related to the github issue you posted? Do you have any extra error logs which are the same as in the report? Otherwise you may have to create a separate issue in github with your error stack trace, as we are not penpot internal experts here.

The last FAIL is the root cause then, OpenID requires the app to reach the OpenID provider, which is on the dashboard domain on Cloudron. Other OpenID enabled apps will also fail then to log in. If you use Cloudflare, have you tried the suggestion? Otherwise check https://docs.cloudron.io/troubleshooting/#hairpin-nat

Thus far is working well. I was able to copy all my work from Public Instance into Cloudron. Will keep yous posted on any issues. Thanks again!

Hi @nebulon, I've tested the new package. Login is now possible for invited users, it's perfect. And editor role is now correctly applied, I don't know if it was a mistake from me at first, but it's now solved. Thanks a lot for assistance !

Not sure, I can't reproduce this issue here, neither on a fresh nor older instance. Have you found anything related in the upstream issue tracker? Given the exception about incomplete user info, is there anything missing strange in the profile? Maybe special characters which penpot potentially can't handle well?

Thanks @nebulon for the prompt fix!

OK again I asked too quickly. You must be logged out for this to work. I was testing sending an invite to myself before I sent them to the rest of the team... All good!

OK, I bumped the RAM a bit

AFAIK, there isn't one - https://github.com/penpot/penpot/discussions/2508 and https://old.reddit.com/r/selfhosted/comments/n63e71/penpot_admin_user/ I see a repo at https://github.com/penpot/penpot-admin but this is experimental