Runs and what happens in one

Overview

A run pulls fresh research for each market you chose, waits for every one of those storefronts to land, and only then drafts anything. A market that has not produced rows yet is still in progress — it is not skipped. You can watch it from the Agent page or leave and come back.

Proposals appear as each market finishes rather than all at the end.

Starting a run

Start a run from the agent page once setup is complete. The run is split into durable phases rather than one long request, so it survives you closing the tab: it continues in the background and picks up where it was.

A run may retry if something fails, within a bounded number of attempts, and a stale run is eventually advanced or failed rather than left hanging.

Progress

A running agent shows a checklist, and ticks each line as it finishes.

• Reading your store metadata
• Loading your keyword research
• Comparing against competitors
• Finding keyword gaps
• Ranking opportunities
• Drafting metadata suggestions
• Estimating search-visibility gain

If you granted the screenshots scope, a final Rendering localized screenshotsline appears once that work starts. The line in progress follows the live Product phase — waiting for each market's Radar research, then drafting, then packing the keyword field — so a long step still shows movement.

The first run for an app is the slowest, because nothing is cached yet and every market needs a full pull. You can leave the page or reload it: the run continues in the background and the checklist picks up where it was.

Generate suggestions

Generate suggestionsopens the same market picker as setup. Choose up to five storefronts — ones you already run, new ones, or a mix — then it starts a full pass on that set. It remounts today's App Store ranks, builds the keyword plan without your previous protect, exclude, or track overlay, and redrafts every scope you allowed. The full-screen progress checklist stays up until this pass finishes; a reload while it is still running brings that screen back.

It is not the same as the weekly recurring pass, which works from cached data. Use it when something changed that the cache would not know about: a new version, a repositioned app, a market you have not looked at in a while.

Recurring work

Once an agent has been started, its metadata pass recurs weekly. A daily check decides whether this week's pass is due.

Weekly is a fixed product behaviour rather than a cadence you configure.

After approved metadata reaches the store, Spek keeps that version stable for a seven-day measurement window. The Agent page shows the remaining days and hours. When the window closes, the next eligible pass checks for stronger keyword opportunities; it may still decide that your current metadata is best.

The recurring pass produces proposals exactly like a manual run does, and like a manual run it never publishes anything to a store. Its output waits in the inbox until you look at it. See what the agent will never do.

Common questions