Ray activity and runs

Every question you ask Ray is a run: the model turns it into a series of tool calls, and the tool calls turn into an answer. Relay keeps a record of each one, whether you asked in the workspace chat bar or in Slack, and shows them in one place.

Open Settings → Plan & usage → Usage, or type Show Ray Runs in the command palette (⌘K, or press g then r).

Ray activity

The Ray activity section is the summary:

Runs

How many runs in the last 24 hours and the last 30 days, across the chat bar and Slack.

Estimated cost

What those runs cost in model usage, estimated from the published per-token rates for the model Ray used. This is an estimate for your own reference, not a bill.

Tokens

Everything Ray read and wrote for you, including the cached part of each request. Cached tokens are what keep a long conversation from costing more each turn.

AI prompts

The row in the allowance ledger above counts prompts against your plan’s daily limit. Ray activity is what those prompts did.

Recent runs

Below the summary is a list of your most recent runs, newest first. Each row shows its status, where it came from (App or Slack), when it started, its estimated cost, and the first line of what you asked.

Stop appears on a run that is queued or still running. A queued run stops at once. A running one finishes the step it is on and then stops, and the row reads Stopping until it does. This works from any tab, and for Slack runs from your desk: a question you asked from your phone can be stopped here.
The transcript opens when you expand a row: the prompt in full, each tool Ray ran and whether it succeeded, the reply, and a footer with the step count, tool calls, tokens and cost. While a run is still going the transcript refreshes on its own.
Done, Failed and Cancelled say how a run ended. A failed run keeps its error message in the row.