Digital Surface

Digital Surface captures and organizes the scattered information from your workday into a searchable personal archive that AI agents can actually access. It records meetings, transcribes dictation, preserves screen history through OCR, and lets you query everything through a Claude powered chat interface. The core problem it addresses is that valuable context gets lost constantly throughout knowledge work, whether it's something mentioned in a call three weeks ago that's suddenly relevant, a terminal command you ran but can't remember, or decisions made in conversations that never got documented anywhere. This tool builds a persistent record of your digital activity that becomes genuinely useful for recall and AI assisted work.

The system operates through four capture mechanisms that run on your Mac. Meeting transcription happens automatically without joining calls as a participant, which means it works with whatever video conferencing tool you're using without needing integrations or bot permissions. Transcription happens locally using Whisper or Apple's Speech framework, producing searchable markdown and JSON files that show who said what and what was decided. You end up with timestamped, speaker attributed records of every call without manually taking notes or asking people to repeat themselves.

Dictation works by holding the Globe key and speaking, with transcription handled entirely on your machine. The text lands directly in whatever text field is active at the time, whether that's a notes app, an email draft, a terminal, or a text editor. There are no app specific integrations to configure because it works at the input level. You can speak a thought and have it appear wherever your cursor is, which works surprisingly well for quick capture when typing feels like friction.

Screen history performs periodic OCR to extract text from visible screens while discarding the raw images themselves. Only timestamped text content and app names get retained, which keeps storage reasonable while still enabling searches like finding your terminal history from weeks prior or locating that configuration value you remember seeing in a browser tab. This approach balances utility against privacy concerns since there are no screenshots sitting on disk, just extracted text tied to timestamps and application context.

Everything stays encrypted with AES-256-GCM per user. The system redacts secrets before storage and excludes password managers from capture entirely. Raw audio files remain on your local machine until transcription completes successfully, only then getting uploaded if cloud sync is enabled. When you want to find something, the chat interface queries your captured files using Anthropic's Claude, citing sources so you can see exactly which meeting transcript or screen capture provided each answer. This citation approach builds trust since you can verify what the AI tells you against the original record.

One particularly useful feature involves agent integration through MCP. Your own Claude instance can access these files via an MCP server, enabling personal AI agents to analyze your work history without tokens counting against Digital Surface's usage limits. This transforms the product from a search tool into infrastructure that makes your AI assistants genuinely aware of what you've been working on. When you ask your agent about a project, it can pull from meeting notes, screen captures, and dictated thoughts rather than operating with no context about your actual work.

The Starter plan costs twenty dollars monthly and includes five dollars of model credits, enough for roughly thirty to one hundred chats or similar activity levels, plus ten gigabytes of storage. The Pro plan at two hundred dollars monthly provides fifty dollars in credits and one terabyte of storage for heavier usage or longer retention. Both include the desktop capture application, and overage credits are available at cost plus ten percent when monthly allocations run out. Pages, history, and connected sources remain unlimited across both tiers.

Engineers, designers, and anyone managing complex workflows where historical context matters will get the most value here. The product positions itself for knowledge workers who need to recall decisions, conversations, and screen activity from days or weeks prior without relying on memory or manual note taking. A twenty five minute onboarding call with a founder helps configure capture settings for your specific workflow and points your agents at the captured surface.