Understanding Scopes & Permissions
Scopes define exactly what a connected service allows your agent to see or do. Instead of granting blanket access to a tool, Gatekeeper works scope-by-scope — so a connection can be as narrow or as broad as you choose.
How scopes work
When you connect a service, you'll see a list of specific, individually-approvable permissions. For example, connecting Slack might offer scopes like:
- Read channel messages
- Send messages
- Read user profiles
- Manage channels
You don't have to approve all of them. Approving only "Read channel messages" means your agent can see conversations but can't post, invite users, or change settings — even if it tries.
Principle of least privilege
The safest setup is to grant only the scopes your agent's task actually requires, and nothing more. If you're not sure whether a scope is needed, start without it — you can always add it later if your agent hits a wall it genuinely needs to get past.
Updating scopes later
Business needs change, and so can your scopes:
- Go to Services and select the connected service.
- Choose Reauthenticate or Edit permissions.
- Adjust the scopes and confirm.
This doesn't require disconnecting existing policies — your policies will simply apply to whatever the new scope set allows.
How scopes relate to policies
Scopes set the outer boundary of what's possible. Policies then decide what's allowed within that boundary. For example, a service might have write access scoped in, but a policy can still block or require approval for specific write actions. Scopes and policies work together — scopes control access to the tool, policies control behavior within it.