What AI is allowed to do
AI tools sign in as you — only your permissions, only your company's data
An AI tool connected to AxHub doesn't use some separate admin account — it signs in as you and can only use the permissions you already have. So even when you delegate to AI, anything you can't see in the console, the AI can't see either.
How sign-in works
On first connect a browser opens; sign in with your usual AxHub account, consent, and you're done. It's standard OAuth, so there's no client_id or secret to create by hand — registration and token refresh are handled by the client.
- You sign in once. After that, it continues automatically.
- If a token expires or becomes invalid, the client prompts you to sign in again.
What's allowed
Only your company's data is visible
One session is bound to one company (tenant). Which company that is comes from your signed-in membership — data outside your company never even appears in listings.
| Your state | Behavior |
|---|---|
| Member of one tenant | Connected to that tenant automatically |
| Member of several tenants | Pick a tenant at login |
| No tenant membership | No access |
| Membership revoked after connect | Blocked at the next refresh |
An AI tool can't point at a different company either — tools take no tenant argument at all, and the server determines the tenant solely from the verified login token.
Permission scopes
Allowed access comes in two tiers.
| Scope | Covers |
|---|---|
data:read | The default and minimum — every read tool: app, deploy, table, and connector reads |
deploy:write | Opens only when the server's deploy-write gate is on — operational tools such as app creation, deploy triggers, and environment variable changes |
- Calling an operational tool with read-only permissions is rejected (
403 insufficient_scope). - Even with the scope, the final boundary is your permissions — every read and change stays within the access you, the signed-in user, already have.