Second Brain

Your cache ships with purpose-built databases: Tasks, People, Decisions, Meeting Notes, Learnings, Projects, Goals and OKRs, Resources — and you can add your own. Each database has typed columns (text, number, single or multi-select, date, relations to other rows), so content behaves predictably every time.

A Task isn't a sticky note. It has a status, an owner, a due date, and a link to the project it belongs to. Your assistant can filter it, sort it, and reason about it the same way you would in a spreadsheet — without you rebuilding the spreadsheet.

Every cache has an Inbox, and it's the first stop for anything new. A half-formed idea, a link from a colleague, a forwarded email, a voice memo transcript — all of it lands here as unprocessed until you decide what it is.

From the Inbox, file each item into the right database yourself or hand it off to Inbox Triage, a built-in blueprint that reads each item, picks the right destination, and moves it for you. Either way, captures stop being a graveyard of one-liners scattered across the cache.

Folders answer where. Links answer why. Two chunks can connect with a labeled relationship — supports, contradicts, expands, summarises, see also, references — so you can trace a decision back to the evidence behind it, or pull every learning that came out of a single client engagement.

Content compounds instead of dead-ending. Every new note has somewhere to attach, and old notes stay reachable through the links pointing into them.

Rows in a database aren't opaque table cells. Each one is a full chunk with its own body text underneath the structured fields — so a Task row can carry status, priority, and due date and hold the running commentary of what you've tried, what's blocked, and what you decided not to do.

That duality is what makes the same content usable in two modes at once: a filterable list when you want the dashboard view, a searchable body when you want context.

Ask in plain language. Search pulls back relevant meeting notes, decision records, and tasks from across every database in the active cache — matching on meaning, not on keywords. A query about "pricing we landed on last quarter" finds the Decisions row even if it never uses the word "pricing".

The same structure is available wherever you connect ContextCache through MCP — Claude Desktop, Cursor, Claude Code, and other supported clients. An assistant can read entries, run search, move chunks between databases, and follow links within the permissions you grant it.

For setup, see MCP Server.

Dashboard overview · Knowledge Layer · System Rules · Automations