Proton's Lumo Desktop Arrives as a Privacy-Focused Alternative to ChatGPT and Claude

Swiss privacy leader Proton has released Lumo Desktop for Windows, macOS, and Linux, bringing a privacy-first desktop AI assistant with local file access, MCP integration, and zero-access encryption to rival ChatGPT and Claude.
Proton, the Swiss privacy technology company behind end-to-end encrypted services like Proton Mail and Proton VPN, is expanding its generative artificial intelligence ecosystem directly to the desktop. In November 2026, the company officially launched Lumo Desktop for Windows, macOS, and Linux, positioning the native client as a privacy-first alternative to the desktop applications offered by OpenAI's ChatGPT and Anthropic's Claude.
As AI assistants evolve from browser-based chatbots into desktop agents capable of reading local directories, parsing inboxes, and querying internal company tools, users and enterprises face growing concerns over how sensitive local data is handled in the cloud. With Lumo Desktop, Proton addresses those concerns head-on by combining local workspace integrations with zero-access encryption and strict permission boundaries.
Deep Integration With Local Files, Proton Mail, and MCP Servers
Rather than acting as a simple web wrapper, Lumo Desktop is built as a full-featured native workspace that connects directly to the tools and files on your machine. Out of the box, the client can index and interact with local folders, connect to encrypted Proton Mail accounts via Proton Mail Bridge, and link with third-party productivity platforms including Jira, Confluence, and Dropbox.
For software engineers and technical power users, Lumo Desktop also ships with native support for the Model Context Protocol (MCP). Developers can wire custom MCP servers directly into the desktop client, allowing Lumo to query local databases, development environments, and internal APIs without exposing proprietary infrastructure to unverified cloud pipelines.
Granular Control With a Tri-State Permission System
Giving an autonomous AI agent access to local directories and enterprise tools naturally introduces operational risk. To prevent unintended actions, Proton built a tri-state permission architecture into Lumo Desktop that governs every connected tool and file path:
- Allow Automatically: Permits low-risk, read-only operations—such as summarizing a folder of PDFs or checking calendar availability—to run seamlessly in the background.
- Require Manual Approval: Prompts the user for explicit confirmation before executing state-changing actions, such as modifying a local file, drafting an email, or creating a Jira ticket.
- Block Entirely: Strictly revokes the AI assistant's access to designated folders, connectors, or external integrations.
Automating Workflows With 'Schedules' and 'Skills'
Alongside local connectors, the desktop release introduces two major workflow automation modules designed for daily productivity:
Privacy Scandal at Claude AI: Shared Chat Links Were Indexed in Google Search
Anthropic Announces Claude Fable 5.1 and Mythos 5.1: Smarter and 45 Percent Cheaper
- Schedules: Users can configure recurring automated tasks that run at specified times—such as compiling a morning briefing from overnight Proton Mail threads and Jira updates. Crucially, each scheduled job operates inside its own independent permission boundary, ensuring unattended tasks only touch explicitly authorized data sources.
- Skills: A library system where users can save reusable instructions, formatting templates, coding standards, and reference files that can be invoked across projects with a single click.
Zero-Access Encryption and Swiss Data Sovereignty
What truly separates Lumo Desktop from US-based frontier AI providers is Proton's foundational security model. Saved chat histories in Lumo are protected by zero-access encryption, meaning encryption keys remain under the user's control and not even Proton can inspect stored conversations.
Furthermore, Proton guarantees that user prompts, local file contents, and connected app data are never used to train AI models or build advertising profiles. All processing and metadata storage take place on European infrastructure headquartered in Switzerland, keeping user data outside the legal reach of statutes like the US CLOUD Act. In keeping with Proton's open-source tradition, the source code for Lumo Desktop is publicly inspectable so independent security researchers can verify its privacy claims.
Who Should Consider Switching to Lumo Desktop?
For legal professionals, financial analysts, journalists, healthcare workers, and software engineers bound by strict non-disclosure agreements (NDAs) or European data compliance mandates, Lumo Desktop delivers the convenience of modern agentic desktop workflows—complete with MCP extensibility and scheduled automation—without sacrificing confidentiality.
This week’s posts, in your inbox every Monday
A short summary of the posts published that week on AI, Claude Code, MCP and web development. No new posts, no email.
- One email a week
- No ads
- One-click unsubscribe








