Sanctum: Writing & Research for Mac

Sanctum: Writing & Research, developed by Spencer Tally, is a privacy-focused Mac application that bridges deep research and long-form writing. It lets users collect PDFs, web articles, and text files and then use the integrated Ember assistant to synthesize imported material into drafts and answer document-specific queries. The app stores work as local Markdown files, supports research boards and manuscript compilation with citations, and targets writers, academics, students, and journalists who need controlled, centralized workflows.

How does Ember integrate with imported research?

Ember operates as a bring-your-own-key assistant, sending prompts directly to external AI providers using the API key you supply. That design means the app itself does not broker your prompts through developer servers; queries that reference imported documents are forwarded from the desktop to the chosen model. Network access is required for those AI calls, while other editor functions continue to run without connectivity.

How accurate are AI-generated drafts compared to manual research?

The assistant produces outputs that reflect the selected external model and the quality of the source material you import. Syntheses and answers draw on patterns in the model's training plus the content extracted from your files, so factual reliability depends on prompt clarity, source completeness, and the chosen provider. Users should independently verify important factual, legal, or medical claims rather than treating drafts as authoritative.

What file types and workflows does it accept for research intake?

Sanctum imports common research formats including PDFs, web articles, and raw text, then exposes those items on structured research boards. Files and notes are stored as plain Markdown and standard file types inside a vault folder you select. You can synchronize that folder by placing it in a cloud-synced directory such as iCloud Drive, Dropbox, or Google Drive, while only AI queries and external fetches require an online connection.

Does it protect research privacy and avoid application lock‑in?

The app favors local control: no user accounts, no analytics collection, and no proprietary cloud servers are part of its architecture. Documents live as editable plain files, which reduces lock-in and keeps your material accessible outside the tool. The trade-off is that AI processing is handled by third-party providers, so privacy of prompts depends on the external service terms and the API key you supply to the assistant.

A focused choice for individuals who prioritize control over convenience

Sanctum suits writers and researchers who want strict local control over source files and prefer to route AI calls through their own API credentials. It assigns responsibility for AI model selection and cross-device syncing to the user, so teams seeking an integrated cloud sync or an included hosted AI service may find its model less convenient. For solitary authors who value file portability and data sovereignty, it is a practical option.