Search documentation
Search documentation

Administration · Concept

Protect workspace data during analysis, export, and AI-agent connection.

Updated: Aug 27, 2026

How to work safely with data and access

Safe work begins with minimum permissions, careful filters, and never placing secrets in the interface, documents, or messages.

Use minimum permissions

Give a person or agent access only to the required workspace, team, and dataset. A report link must not substitute for configuring a role.

Check exports

Before sending a file, verify period, fields, and recipient. Do not send personal or restricted data through insecure channels.

Protect tokens

Provide tokens and secrets only through protected client settings. Do not put them in URLs, prompts, comments, screenshots, or logs.

Use AI agents safely

Before an analytics query, an agent should verify available workspaces. Do not ask it to bypass row or column restrictions or guess object identifiers.

Next step

Read connecting an AI agent with MCP when programmatic access to authorized data is required.

Was this article helpful?