Plan execution on Grafana Labs

Separate configuration access from results

Sat, 17 Oct 2026 02:03:35 +0100

Who needs the checkout summary?

The saved automation and the conversation produced by a run have different access rules. Visibility controls access to the saved configuration.

Personal automations appear as Just me to their creator. Assistant administrators with the appropriate permissions can also view and edit that configuration. Everybody makes the configuration available to other users with the right access.

For user-mode runs, sharing the configuration doesn’t automatically share the resulting conversation. The user who triggered the run must explicitly share the chat for others to read it.

Personal visibility is a useful starting point while the checkout prompt is being tested. A team may later need the configuration, the results, or both.

Understand whose permissions a run uses

Sat, 17 Oct 2026 02:03:35 +0100

A test should represent the scheduled run

The checkout summary can only use resources available to the identity executing it. For an automation configured to run as User, the trigger determines whose access and usage apply:

flowchart LR
    accTitle: Execution identity for automations configured to run as User
    accDescr: A manual run uses the permissions and Assistant usage of the person who starts it. A scheduled run uses the automation creator's permissions and Assistant usage.
    subgraph U["Run as: User"]
        M["Manual run"] --> I["Person who starts it"]
        I --> MP["Their permissions
and Assistant usage"]
        S["Scheduled run"] --> C["Automation creator"]
        C --> CP["Creator's permissions
and Assistant usage"]
    end

Why the creator’s test matters

A teammate’s successful manual run doesn’t prove that the creator’s scheduled run can access the same data. When the creator tests the prompt, both triggers use the same identity.

Access can change after testing. A summary that previously worked can fail if its creator loses access to a required data source.

Match the schedule to the need

Sat, 17 Oct 2026 02:03:35 +0100

When will you read the summary?

A checkout summary read on weekday mornings may need five scheduled runs in a normal week. Running it more often adds Assistant usage even when nobody needs another result. Manual tests also consume usage.

The schedule’s timezone determines which local time “9 a.m.” means. An explicit timezone such as America/New_York or UTC makes the intended timing clear.

Automations require at least 15 minutes between scheduled runs. Token usage depends on the work each run performs, so run count helps plan frequency but isn’t a fixed price estimate.

Notifications are optional. Results remain available in Assistant conversations without external delivery.

Resources

Sat, 17 Oct 2026 02:03:35 +0100

Resources

Review access rules, schedule behavior, and usage for the planned automation.

Documentation

Choose visibility explains configuration and conversation access.
Understand how runs behave covers execution identities and usage.
Manage Assistant access describes the permissions needed for automations.
Schedule format explains timing and timezones.
Assistant pricing and usage explains how recurring work consumes usage.