Skip to main content
Pro and Team include a dedicated database for your workspace. Open Data workspace in the console to create tables, edit records, and build dashboards. Your tables persist independently of the raw events in project datasets. The workspace opens with room for your data. Choose Browse data to open the table, dashboard and Workspace Brain navigation. Hide sidebar restores the full width; your choice stays in place as you switch between these views.

Create a table

Choose Browse data, then Create table, give it a name, and add fields. Fields support text, numbers, booleans, dates, select options, and JSON. A field’s label is what people see; its key is what data mappings, workflows, and API clients use. Use Add record to enter data or click a cell to edit its record. Changes use record versions, so an outdated edit returns a conflict instead of overwriting someone else’s work. You can change field labels and add fields or select options. Existing keys and types stay stable for automation.

Save views and build dashboards

Click New view to choose visible fields, reorder columns, add filters, and sort records. All filters must match. Sorting and filtering cover the entire table, including records on other pages. Edit view changes a saved view. Choose Browse data, then Create dashboard. Add widgets for record counts, sums, averages, or grouped bar charts. Choose the table for each widget and optionally select a saved view to filter its data. Grouped charts show the top 20 groups. Use Refresh to recalculate the dashboard.

Connect webhooks and data sources

  1. Configure a webhook or data source to populate a project dataset.
  2. Open the table and choose Connect data & workflows → Connect dataset.
  3. Select the project and dataset. Map each destination field to an incoming path, such as customer.email. Leave a mapping empty for an empty field.
  4. Preview a sample JSON payload, then save the connection.
New matching dataset records populate the table automatically. Workflow and AI output that is delivered into the ingestion pipeline can use the same connection. Numeric strings convert to numbers, and "true" / "false" convert to booleans. Other invalid values appear as a connection error. Import existing records processes up to 100 source records at a time. Continue with Import next 100 when more records remain. Imports are safe to repeat: they preserve existing table records, human edits, and deletions. Correct mapping errors and replay while the original records are still within dataset retention.

Use records in workflows and agents

On a data connection, enable Allow this project’s workflows to read and write this table. A project’s credentials cannot use its workflows to access a table without this grant. In the workflow builder, add a Table records step:
  • Read records: choose a table, optional saved view, and a limit of 1–100. The default result is available at table_result.rows. Place an AI or Agent call after this step to use the records as context.
  • Insert record: map field keys to values or workflow placeholders, for example {"summary": "{{analysis.summary}}"}. Retried steps do not create duplicates.
Connected agents can use list_workspace_tables, query_workspace_table, insert_workspace_record, and update_workspace_record. These tools require a workspace-scoped credential and honor its read/write grant. Updating a record requires the current version and the complete desired values.

Limits and lifecycle

Each record supports up to 64 KiB of serialized values. The underlying database’s storage limit also applies. Workflow results retain the platform’s existing size limits; read fewer records if a workflow result is too large. Downgrading below Pro pauses dedicated-data operations without deleting the database. Upgrading restores access. Owners and admins can still use Settings → Export & closure to export the data after a downgrade. Workspace closure deletes the dedicated database along with the rest of the workspace. Tables, views, dashboards, and records are included in the workspace JSONL export. A complete export ends with "complete": true; a missing final marker means the download did not finish.

Connect records into agent knowledge

The Workspace Brain adds a workspace map, a spreadsheet, visual relationships, bounded agent context and execution lineage on these same records. Choose which tables and individual records belong in the Brain without deleting the source data. Agent and workflow mutations consolidate asynchronously, and conflicting proposals can be reviewed before applying them.