Table projects: structured data for AI

Memocore is memory for AI. Usually it holds memos: notes, decisions, agreements. Your agent finds them by meaning and answers.

But some knowledge isn't text, it's records: leads, deals, orders, candidates. You don't need to "remember" them, you need to count and filter them. "How many new leads this week?", "Deal total by month" — search by meaning doesn't help here.

Table projects

When creating a project, you can now choose the "table" type. Every memo in it is a row, and rows have columns with types:

• text;
• number;
• date and time;
• checkbox;
• single choice;
• multiple choice.

Every column has a description — what it holds and how to fill it in. The agent reads it and fills in the rows correctly.

A column can be made required or unique: the agent won't create a duplicate, it will update the existing row.

The table on the site

On the project page, rows are shown as a familiar table. Columns can be added, edited and reordered right on the site. Clicking a row opens it in full.

The rows themselves are added by the agent: just tell it what to record.

Exact queries

The agent can filter, sort and count rows:

• "Show leads with status new and an amount over 1000, newest first"
• "How many orders were there in September?"
• "Deal total by month and by manager"

Available: count, sum, average, minimum and maximum, grouped by columns and by day, week, month or year.

How to create one

Ask your agent: "Create a table for leads: name, email, status, amount, date." It will pick the column types and write their descriptions itself. Or choose the "table" type when creating a project on the site.

The project type and the column types can't be changed later, so it's best to think the structure through right away.

Table projects are available in company projects (Team plan).