Mobile companion for Lumo Desktop: remote permission approvals and remote task start
Summary
The Lumo desktop app can now use MCP servers and create files on my computer. That makes it useful for real work, but only while I’m sitting at the computer. I’d like the Lumo mobile app to control the desktop app, so I can start tasks and handle permission prompts while I’m away.
- Approve or deny permission requests from the mobile app When Lumo Desktop needs permission for an action (an MCP tool call, writing a file and so on), it currently waits until I respond at the computer. Long tasks often stop for hours because of one prompt.
Requested behavior:
A push notification on my phone when the desktop app needs a permission
The notification or the app shows what Lumo wants to do: tool or server, action, target file or resource
I can allow once, allow for this session, or deny it from the phone
The desktop task continues right after I respond
- Start desktop tasks from the mobile app I’d like to give Lumo a task on my phone and have it run on my computer, with access to my configured MCP servers and local files.
Requested behavior:
The mobile app lists my linked desktop devices and shows which ones are online
I enter a task on mobile, and it runs in the desktop app on the device I pick
I can follow progress, get a notification when it finishes, and see or download the results on my phone
Permission requests from these tasks go to mobile, as in point 1
Reference
Other assistants already do this. Claude and ChatGPT both let you approve permissions and start tasks on a linked computer from their mobile apps. Lumo would need this to stay competitive for agent-style work.
-
Alexander Simbun commented
Strong +1, and I can add a concrete failure case from this week that shows exactly why both halves of this matter.
I've been using Lumo Desktop daily to manage my local mailbox through Apple Mail (reading, searching, filing, spam triage — it works beautifully). Yesterday morning from my phone, I asked Lumo Mobile for my latest email. It had no route to the desktop instance and quietly answered from the previous session's memory instead — the reply looked current but wasn't. That's the subtle trust problem this proposal solves: it's not just about starting tasks remotely, it's about the mobile app knowing what it can and cannot currently see.
So beyond what's already proposed here, three design details I'd fold in from that experience:
Always-visible presence state — your device list with online status is the right idea; I'd extend it so every mobile answer generated via the desktop is labeled as a live pull versus cloud-only, and offline means saying so explicitly rather than serving stale answers.
Tiered approvals — not all permission prompts are equal. Read/search (list messages, find files) shouldn't need a phone tap every time; modifies (move, rename) could be session-scoped; only destructive or outbound actions (delete, send, empty trash) should require explicit tap-to-confirm. Otherwise the approval flow becomes the bottleneck it's meant to remove.
Zero-access relay — task content and results traveling phone ↔ desktop should be E2E encrypted with the relay server learning nothing. That's Proton's home turf and would differentiate this from Claude/ChatGPT's implementations.
For what it's worth, this pairs naturally with the "Lumo integration into other Proton apps" idea also in this forum — cross-app reach plus cross-device control is a coherent "Lumo everywhere" story, and the same people will likely want both.