Skip to content
Glowo
Esc
↑↓navigate↵open⌘Jpreview
On this page

Scopes and consent

Exactly what you are granting an AI agent, and how to take it back.

The scopes

Each scope an agent can hold, in the words the consent screen uses, and which of the three presets carries it.

ScopeGrantsRead OnlyOperateFull Access
monitors:readRead monitors and heartbeats✓✓✓
monitors:writeCreate, update, pause and resume monitors; run on-demand checks—✓✓
monitors:deleteDelete monitors——✓
status_pages:readRead status pages✓✓✓
status_pages:writeCreate, update and publish status pages——✓
status_pages:deleteDelete status pages——✓
incidents:readRead incidents✓✓✓
incidents:writeFile, update, resolve and reopen incidents—✓✓
incidents:deleteDelete incidents——✓
postmortems:readRead postmortems✓✓✓
postmortems:writeWrite and publish postmortems—✓✓
postmortems:deleteDelete postmortems——✓
notifications:readRead notification channels and delivery history—✓✓
notifications:writeManage notification channels and send tests—✓✓
notifications:deleteDelete notification channels——✓
maintenance:writeSchedule maintenance windows—✓✓
identity:readView workspace and plan✓✓✓

Deleting is a scope of its own for every kind of resource. Operate carries none of them: it is the preset people approve without reading, so it can change things but never destroy them.

Presets

Preset Good for
Read Only Summarising, investigating, reporting
Operate Day-to-day on-call: create and pause monitors, run checks, acknowledge, open, update and resolve incidents, write and publish postmortems, manage channels, schedule maintenance
Full Access An agent you have come to trust: it can also set up and publish status pages, and delete monitors, incidents, postmortems, status pages and channels

Read Only is the right starting point, and it is more useful than it sounds — most of what people want from an agent is “tell me what is wrong”, which needs no write access at all.

What no scope can grant

Not available to any agentWhat it covers
workspace:manageManage workspace settings
billing:manageManage billing and subscription
members:manageInvite and manage members
audit:readRead the audit log

These are session-only. An agent can never change your subscription, add members, alter workspace settings, or read the log of its own activity.

When an agent signs in with OAuth, you are shown a consent screen naming the app, who stands behind it, where your approval goes, the workspace it will act in, and the levels above in plain language. Nothing is granted until you approve, and every sign-in shows the screen again, even for an app you approved before. Tools shows which level each tool needs.

Who is asking

An app chooses its own name, so the line under it says who stands behind it:

  • Published by a site, with a badge: the app’s sign-in details are published on that site, and Glowo knows the site as the app’s vendor (claude.ai for Claude Code, vscode.dev for VS Code, chatgpt.com for Codex).
  • Published by a site Glowo does not know: the name is the app’s claim, and the address is what to check.
  • Named by the app itself: the app registered itself, so nothing confirms its name, which is shown in quotes.

Where the approval goes

  • An app on this computer, with its port, or the app that opens a kind of link (for example cursor: links): any program on your computer could be the one receiving it, so approve only a sign-in you just started.
  • A web site: the site’s address, marked Verified only when a known publisher’s sign-in returns to that publisher’s own site over HTTPS.

The workspace

The approval binds the workspace the screen shows, and only that one. If another tab switches your session to a different workspace before you approve, the approval is refused, and reloading the screen shows the new workspace.

You choose the preset, not the client. What a client asks for is a ceiling for that authorization attempt, not a grant. Codex, for one, asks for every scope Glowo publishes; the screen still opens on Read Only, and what you approve there is what the agent gets. A later attempt can ask for a different set, so the ceiling is per-connection, not permanent.

Grants are workspace-scoped: approving an agent in one workspace gives it no access to another. An app that signs in on several machines (Claude Code on your laptop and your desktop, say) is one connection per workspace, and revoking it signs out every machine.

Reviewing and revoking

Connected agents are listed in the dashboard under Settings → MCP, and over the API:

GET  /mcp/grants
POST /mcp/grants/revoke-batch

Revocation takes effect immediately.

Advice

  • Grant read-only first, and widen only when you have a concrete reason.
  • One grant per client. Separate grants for your editor and your terminal agent mean you can revoke one without disturbing the other.
  • Check the audit log after the first few sessions to see what the agent actually did.

Was this page helpful?