Proton brings privacy-focused Lumo AI to Windows, macOS, and Linux
Proton is expanding its privacy-first AI assistant, Lumo, beyond the web and onto personal computers. The company has launched Lumo Desktop for Windows, macOS, and Linux, giving users a way to use AI for local tasks while keeping tighter control over personal data, files, and app permissions.
The move comes as AI assistants are quickly evolving from simple chatbots into more capable AI agents that can browse information, interact with apps, summarize files, automate workflows, and complete tasks on a user’s behalf. While that can be convenient, it also raises a major question: how much access should an AI tool have to your computer?
Proton’s answer is to make access more transparent and customizable. Lumo Desktop is designed to work with local files, folders, connected apps, and automation tools, but users can decide exactly what the AI is allowed to do. Some actions can run automatically, others can require approval, and certain capabilities can be blocked completely.
That permission-based approach is one of the main selling points of Lumo Desktop. Many AI desktop apps need broad system access to be useful, which can increase privacy risks. Proton is positioning Lumo as an alternative for people who want AI productivity features without handing over unnecessary control of their personal information.
The company says Lumo conversations, connected app data, and personal files are not used for targeted advertising or to train AI models. Proton also says chat history is encrypted so that only the user can read it. According to the company, it does not keep server-side logs, and the limited data it does store is protected under Swiss privacy laws.
Lumo Desktop includes several connectors that allow the AI assistant to interact with different tools and services. These include local file-system access, shell execution, browser content fetching, Proton’s own email service, and third-party platforms such as Confluence, Dropbox, and Jira.
The app also supports local and remote Model Context Protocol servers, often called MCP servers. These are used by AI agents to access external data sources and expand what they can do across different workflows.
For each connector, users can choose how much freedom Lumo has. For example, the AI may be allowed to read certain files automatically, but require approval before editing or running commands. Users can also disable individual tools directly from the Tools menu inside a conversation.
Lumo Desktop also supports scheduled tasks. This means users can create recurring automations with their own permission settings. One task might be allowed to run without interruption, while another may need approval before accessing files or sending information to a connected app.
Another key feature is Skills, which allow users to save reusable instructions, templates, and reference files. This can help Lumo remember preferred formats, writing styles, project details, or repeated workflows across conversations. Proton says an earlier version of this feature, known as Custom Lumos, will be moved into the new Skills format.
Behind the scenes, Proton says it is investing in its own AI infrastructure. The company’s first NVIDIA B300 cluster, built using Open Compute Project standards, is now going live in Switzerland. Proton says its data is hosted on European infrastructure, outside the reach of certain U.S. cloud data access rules.
With Lumo Desktop, Proton is aiming at users who want AI assistance for everyday productivity but do not want to sacrifice privacy. The app brings together file access, app connectors, automation, scheduled tasks, and reusable AI skills, while giving users more control over what the assistant can see and do.
As AI agents become more common on personal computers, privacy and permission management are likely to become major deciding factors for users. Proton is betting that many people want the benefits of AI automation, but with stronger safeguards around personal data.






