Access model

/admin/permissions — labeled Permissions in the sidebar. The rules every tenant works within; nothing on this page gives anyone access by itself, it only sets what tenants are allowed to hand out.

Table of contents

  1. The three-layer model, in one minute
  2. Tab 1 — Permissions and tiers
  3. Tab 2 — Built-in roles
  4. Tab 3 — Tenant guardrails
  5. Tab 4 — Wide grants
  6. Tab 5 — Advanced

The three-layer model, in one minute

A permission is one thing that can be done (e.g. reading data). A role is a named bundle of permissions. A grant gives a role to a key or a person at a scope: one database, all of a tenant’s databases, or the whole platform. Permissions and roles say what; scope says where — so the same role, even Owner, is a safe everyday choice on one tenant’s own databases and a very different decision on the whole platform.

  1. Which databases? — set on each key, checked on every request; a tenant’s key can never reach another tenant’s database.
  2. Which actions? — the role’s permissions (this page).
  3. Which rows? — an optional row policy narrowing a role further, per tenant (see Features → Row policies).

The page has five tabs: Permissions and tiers, Built-in roles, Tenant guardrails, Wide grants, and Advanced.

Tab 1 — Permissions and tiers

Every permission the platform knows about, in one table: what it lets someone do, its tier, who may hand it out, what could go wrong, and which roles currently hold it (computed live from what’s actually enforced, so this table can never disagree with reality).

Tier Meaning Who may grant
1–2 Everyday / removal actions (read, create, update, delete) Tenants, through whichever roles you allow
3 Sensitive (e.g. seeing structure, generating configuration, running scripts) Needs your approval, per tenant
4 *.* — everything, including anything added later Needs your approval; a platform-scoped grant needs platform-only approval

A standing banner on this tab calls out *.* specifically: it grants total access — every action on every table, plus every database action, including anything added in the future. For everyday full access, the page recommends Editor instead; reserve *.* (the Owner role) for a tenant you trust, on that tenant’s own databases, or on the platform for your own operators.

Permissions created by hand for a specific table (rather than the five standard actions) are flagged and summarized here, with the full list on the Advanced tab.

Tab 2 — Built-in roles

The same five roles described in Concepts → Authentication & credentials (Reader, Writer, Editor, Database admin, Owner), shown here as an exact matrix — ten columns: Read, Add, Change, Delete, See structure, Config 🛑, Starter data 🛑, Scripts 🛑, Everything 🛑, Tier — plus how many keys and people currently hold each one. Built-in roles are locked; you can’t edit what they hold, only whether a tenant is allowed to hand one out (Tab 3).

Below the matrix, each role gets its own card with a plain-language sentence of exactly what it allows — this is the same wording a tenant sees when deciding which role to give a key. If any tenant has built its own custom role (Tab 3 → “Roles a tenant may build itself”), those appear in a second table underneath, each labeled with its owning tenant.

Tab 3 — Tenant guardrails

This is where you actually decide what tenants may do with their own workspace, and it’s built around one default that applies to every tenant, plus exceptions for the few that need something different.

Default for every tenant — a form with three independent choices:

  1. Roles a tenant may hand out — a checkbox per built-in role. Reader/Writer/Editor are on by default; Database admin and Owner are off (each marked “needs your approval” until you turn them on), and Owner additionally shows a standing warning explaining it’s full control of that tenant’s own databases.
  2. Where a tenant may grant access — either “Chosen databases only” (default — every key must list its databases) or “Also all of its own databases” (adds an “All my databases” choice when a tenant creates a key; such a key automatically reaches every database that tenant owns, including ones added later, and shows up on the Wide grants tab).
  3. Roles a tenant may build itself — Off (default) or On. When on, the tenant’s own Roles screen gets a “New role” button that can combine read/add/change/delete/see-structure in any mix — never configuration, scripts, starter data, or *.*, regardless of this setting.

A live preview below the form spells out, in plain language, exactly what a tenant will see and be able to do under the choices as currently checked — before you save anything.

Exceptions — tenants that don’t follow the default. Search for a tenant by name or id to give it its own settings (this doesn’t show every tenant, by design — it stays short regardless of how many tenants you have). Each exception can be edited or reverted to “Use the default” at any time.

Changing a guardrail never revokes access a key already has. If you tighten a default or an exception, the save response lists exactly which existing keys now hold more than the tenant is currently allowed to grant — they keep working as-is, but you can see them and decide whether to act.

Tab 4 — Wide grants

Every credential that reaches more than one database of a single tenant — either all of that tenant’s databases, or (for a platform key) the whole platform. This is the review list: narrower, single-database keys never appear here, since they’re not what a guardrail audit needs to focus on.

Each row shows who holds the grant, which tenant (blank for a platform-wide key), what it can do, its reach (tenant-wide or platform), how many times it’s been used, and when it was last used — “never” means exactly that, not missing data. A Review link opens the right keys page to narrow or revoke it. A platform key marked for the whole platform with no tenant attached is flagged as having no clear owner.

If nothing qualifies, the tab simply says so — “No key reaches more than one database of a tenant.”

Tab 5 — Advanced

The raw permission list — every {resource}.{action} string that exists, its description, and which roles hold it. * and database permissions are locked (can’t be deleted from here); anything created by hand can be removed, with a warning that any role holding it loses it immediately, and keys using that role may start getting 403s.

Two creation tools, used rarely:

  • Add permission — a single resource + action (+ optional description).
  • Bulk create — one resource (a real table name) and a checklist of the four standard actions (create/read/update/delete), previewing the exact permission strings before creating them.

Table-specific permissions created here are matched by exact spelling against your database’s real table names — a typo (emplyee instead of employee) silently creates a permission that grants nothing, because no table will ever match it. Prefer letting a role’s table-scoped access be chosen from each database’s own table list elsewhere in the console, rather than typing names here, for exactly this reason.


Copyright © CrudApi. Distributed under an open license.

This site uses Just the Docs, a documentation theme for Jekyll.