Lumo Hits the Desktop, and You Decide What the AI Can Do

Proton has launched Lumo Desktop, a desktop version of its privacy-focused AI assistant, for Windows, macOS and Linux. According to TechCrunch AI, the app goes beyond answering questions. It can work with connected apps, read your personal files and run automations on your computer. Proton’s pitch is that you stay in control of how much access the AI gets.

The timing makes sense. OpenAI, Anthropic, Meta and others already offer desktop apps that can act for you. But giving an AI agent deep access to your machine puts your personal data at greater risk. TechCrunch AI notes that Apple announced tighter security settings this week for apps that request Full Disk Access, which is the kind of access AI desktop apps ask for.

🧭 What Lumo Desktop Brings

  1. Connectors for local and cloud tools: Lumo connects to local tools for file access, shell commands and fetching content from the browser. It also links to Proton’s own email app and to third-party services such as Confluence, Dropbox and Jira. That covers a typical knowledge worker’s day, from documents to tickets.
  2. MCP support, local and remote: The connectors work with both local and remote MCP (Model Context Protocol) servers. MCP is the open standard AI agents use to reach external data sources. Supporting it means Lumo isn’t limited to the services Proton chose to integrate.
  3. Permissions for each connector: This is the key feature. For every connector, you decide which actions Lumo can take on its own, which need your approval and which are blocked. You can also switch off individual tools from the Tools menu during a conversation. Most competitors offer a much blunter on/off choice.
  4. Scheduled tasks with their own permissions: You can set up recurring tasks, and each one gets its own permission settings. A daily inbox summary could run on its own, while anything that sends or deletes still waits for your approval.
  5. Skills for reusable work: Like other AI apps, Lumo supports Skills, which store instructions, templates and reference files the assistant can reuse across conversations. Proton’s earlier Custom Lumos feature will move to the Skills format.

🔐 The Privacy Stack

Proton built its brand on privacy, and Lumo follows the same approach. According to TechCrunch AI, the company says:

  • Chat history is encrypted so that only you can read it. Proton can’t.
  • It keeps no server-side logs.
  • Conversations and data from connected apps are never used for personalized ads or to train AI models.
  • The limited data it does store is protected under Swiss privacy law and hosted on European infrastructure, outside the reach of the U.S. CLOUD Act. That law governs how U.S. law enforcement can get data from cloud providers.

Proton is also building out its own compute. Its first Nvidia B300 cluster, built to Open Compute Project standards, is now going live in Switzerland.

📊 Why This Matters

What stands out here is the shift in what “privacy” means for AI. With a chatbot, privacy mostly meant protecting what you typed. With an agent, it also means limiting what the AI can do: which files it can open, which commands it can run and which messages it can send without asking. Lumo’s approve, auto or block setup addresses that second part directly.

It also puts competitive pressure on bigger players. If users get used to setting permissions per connector, and regulators and platform owners like Apple keep tightening access, broad all-or-nothing permissions will start to look outdated.

⚠️ Caveats Worth Noting

The report doesn’t cover pricing tiers for the desktop app, and it doesn’t say which underlying models power Lumo or how they compare with frontier systems from OpenAI or Anthropic. Privacy only helps if the assistant is also capable enough for real work. The permission controls also depend on users setting them carefully. An agent with shell access set to “automatic” is still an agent with shell access.

🔭 What Comes Next

The interesting thing to watch is whether privacy-first agents can keep up on capability while the big labs move fast. Proton’s new GPU cluster suggests it plans to compete on infrastructure as well as policy. The full details are in the original TechCrunch AI report.

Scroll to Top