System of Record (Data Tables)

Create custom data tables that agents can read and write.

What is the System of Record?

The System of Record (SOR) provides structured data tables that agents can create, read, update, and delete. Unlike the Knowledge Base (which is optimized for unstructured document search), data tables store structured rows and columns — think of them as a simple database your agents can use.

Use Cases

CRM — "Contacts", "Companies", "Deals" tables with consistent column names across agents.
Inventory tracking with item, quantity, and location columns.
Project tracking with task, assignee, due date, and priority.
Research data collection — agents populate rows as they find information.
Configuration tables — dynamic settings agents reference during their work.

Creating a Data Table

Navigate to Data Tables in the sidebar.
Click "New Table".
Name the table (e.g., "Contacts", "Inventory").
Define columns with names, types (text, number, boolean, date), and optional default values.
Click Create.

Column Types

Type	Description	Example Values
Text	Free-form string.	"John Doe", "pending"
Number	Integer or decimal.	42, 3.14
Boolean	True or false.	true, false
Date	ISO 8601 date string.	"2025-06-15"
Select	One of predefined options.	"active", "inactive"

Agent Access

When editing an agent, you can grant access to specific data tables. An agent with access can perform CRUD (Create, Read, Update, Delete) operations on the table's rows during its heartbeat cycle. Access is read-write by default; you can restrict agents to read-only if needed.

Agent Organization

Agents are instructed to stay organized when working with data tables. They follow these rules automatically:

Check for existing tables before creating new ones — agents query the table list and reuse matching tables.
Use canonical CRM names — "Contacts" (not "Leads"), "Companies", "Deals" — with consistent column naming across agents.
Prefer adding rows to existing tables over creating duplicate tables with slightly different schemas.

lightbulb

If multiple agents are working on related data (e.g., sales prospecting + outreach), they will naturally converge on the same "Contacts" table, keeping your data unified.

CSV Auto-Import

When you upload a CSV file (via drag-and-drop or the upload button), it is automatically imported as a new data table. The first row of the CSV becomes the column headers and all subsequent rows are imported as data. The table appears in the Data tab, not the Files panel.

This also applies to agents — when an agent writes a .csv file using the write_workspace_file tool, it is automatically converted into a data table instead of being saved as a raw file.