Proton releases the Lumo Desktop desktop AI application with an emphasis on privacy and MCP support
The company Proton has launched Lumo Desktop for Windows, macOS, and Linux – an AI assistant with local execution, encrypted chat history with no server-side logs, and support for MCP servers. Data is hosted in Europe, outside the reach of the US CLOUD Act.
The company Proton has released Lumo Desktop, a desktop version of its AI application for Windows, macOS, and Linux. According to the company, the application is designed as an alternative to desktop tools from OpenAI, Anthropic, and Meta, with an emphasis on user control over personal data when an AI agent works with connected apps, files, and automations.
Lumo Desktop offers connectors to local tools (file system access, running shell commands, retrieving content from the browser), to Proton's own Proton Mail, and to the external services Confluence, Dropbox, and Jira. The application supports both local and remote MCP servers. For each connector, users can set which actions the AI performs automatically, which require approval, and which are completely blocked; the same permission system also applies to scheduled tasks. The application further supports a Skills feature for storing instructions, templates, and reference files; the earlier Custom Lumos feature will be migrated to the Skills format.
According to Proton, chat history is encrypted in such a way that only the user can read it, the company keeps no server-side logs, and the limited data it does retain is subject to Swiss law. Data is hosted on European infrastructure outside the reach of the US CLOUD Act. The company also states that it is investing in its own infrastructure – according to its statements, the first cluster with NVIDIA B300 chips, built on Open Compute Project standards, is already operating in Switzerland.
Why it matters
Desktop AI agents commonly request broad access to files and the system (Full Disk Access), which increases the risk to personal data – Proton addresses this with granular permission settings per individual action and connector. For companies and individuals dealing with European data jurisdiction or collaborating via Confluence, Jira, or Dropbox, this represents an alternative with a differently configured trust and hosting model compared to US providers.
Two audiences, two different impacts
What this means
For individuals
When choosing a desktop AI assistant, one can now consider a tool with finely adjustable permissions (automatic/with approval/blocked) and data hosting outside the reach of the US CLOUD Act, instead of the broader but less transparent permissions found in competing desktop AI applications.
For a business
Companies using Confluence, Jira, or Dropbox and dealing with requirements for a European data location or avoidance of the CLOUD Act gain an alternative AI tool with a configurable level of automation and action approval per individual connector.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.