Permissions, approvals & audit
How in-app AI stays safe — and what's coming next.
On this page
It runs as the user, every time #
Ask, Analyze, and agents all execute under the signed-in end-user's Tadabase session. There is no elevated "AI account" that sees everything — the AI is bound by exactly the same RLS rules as the person using it.
Per-interface gating #
AI features are enabled per interface and per tool, server-side. You decide which screens get Ask, which agents exist, and which actions each agent may take without approval.
Everything is logged #
Every agent action is recorded with the user, timestamp, the action taken, and the result. AI spend lands in the same usage dashboard as the rest of your account, attributed by source.
On the roadmap #
In-app AI is shipping in stages. Live today: Ask, Analyze, and action agents with approval gates. Rolling out next:
- Saved Ask pages — bookmark a question and re-run it with fresh data.
- Scheduled refresh — re-run a saved question on a schedule and notify on change.
- Richer renders — maps, timelines, and trend lines beyond table and chart.
- Per-user integrations — connect Gmail, Slack, Asana, and 2,700+ apps so agents can act across them.