🛰️ Runs: see everything that happened

Runs answers the question every automation product dodges: “what did it actually do?” Every execution - scheduled jobs, chat replies, email triage, research calls, tool invocations - is a row: when, which agent, what triggered it, how long, which model, and whether it succeeded.

The detail panel is the receipt: open any run for its summary, token usage (input/output/cache), research parameters and result links, tool request/response payloads, and a jump to the related mission. Errors and timeouts show the exact failure - with a “Raise the timeout” shortcut when that’s the fix.

Navigate like a calendar

Day/Week/Month views, ‹ › to move the window (or press ← →), t for today, and the mini-month to jump anywhere in the last four weeks. UPCOMING shows the next scheduled fires with countdowns. Filter by agent or activity type when you’re hunting something specific.

💡 Tip: Weekly 60-second habit: open Runs, glance at the error count and the success rate in the rail. Two errors from the same schedule = fix the schedule, not the symptom - the ⚙ on any scheduled run jumps straight to its definition.

Runs is also your cost microscope: each row shows model + tokens, research rows show credits consumed. When you wonder where a busy day went, the answer is sorted, timestamped, and itemized here.