Sanctum — Your Private, Local AI Assistant

What

Sanctum is a desktop application for running open-source large language models locally on a Mac or Windows device. It is designed for people who want generative AI features without sending prompts, files, or conversation history to a remote service, with local processing and encrypted local storage positioned as core parts of the product.

The product appears suited to privacy-conscious individual users and professionals handling sensitive information, such as personal health, financial, document, coding, or feedback analysis workflows. Its core workflow is to download the app, choose compatible GGUF models, run them on-device, and use them for chat or PDF analysis without requiring an internet connection after setup.

Features

• Local LLM execution — Runs full-featured open-source models directly on the user’s device, reducing reliance on cloud-hosted inference.
• Offline use after download — After installation, the app can operate without an internet connection, which is useful for private or disconnected work environments.
• Encrypted local data vault — Stores chats and related data in a local encrypted repository called Sanctum Vault, helping keep usage data on the device.
• AES-256 protection for stored data — The site states that Sanctum Vault uses AES-256 encryption and is accessible only with the user’s account password.
• Hugging Face GGUF model access — Lets users browse, check compatibility for, and download GGUF models from Hugging Face within the app workflow.
• Private PDF chat and summarization — Supports asking questions about and summarizing PDF files through local processing rather than remote document upload.

Helpful Tips

• Check device suitability early — Local LLM performance depends heavily on available hardware, so teams should validate model size and speed expectations on target Macs or Windows machines before broader rollout.
• Treat privacy claims as device-scoped — The product strongly emphasizes on-device processing and local encryption, but buyers should still review internal endpoint security, password practices, and file access controls.
• Match model choice to the job — Since Sanctum supports downloadable open-source models, practical outcomes will vary by model quality, size, and compatibility rather than by the app alone.
• Prioritize high-sensitivity use cases first — Document review, financial notes, health-related analysis, and internal brainstorming are logical early use cases where local execution offers a clear operational benefit.
• Confirm platform roadmap needs — Current support is listed for macOS 12+ and Windows 10+, with Linux and mobile noted as coming soon, so cross-platform teams should plan conservatively.

OpenClaw Skills

Sanctum could likely work well within the OpenClaw ecosystem as a private local inference layer for document, note, and analyst-oriented workflows. Likely OpenClaw skills around it could include PDF interrogation agents, local research copilots, call-note summarizers, code review assistants, and personal knowledge agents that process sensitive files without routing them through cloud AI services. The page does not mention a native OpenClaw integration, so this should be treated as a likely deployment pattern rather than a confirmed capability.

In practice, combining Sanctum with OpenClaw-style skills could give professions such as finance, healthcare-adjacent operations, legal support, software development, or user research a more controlled way to use AI on confidential material. A likely outcome is a workflow where OpenClaw orchestrates repeatable tasks and decision support, while Sanctum provides local model execution and private document context, shifting AI adoption from generic chat toward secure, role-specific desktop automation.