10 components

Run

One run in detail: each tool call with its live timer and result, waterfalls of where the time went, the terminal it types in, how full the context is and what it all costs.

Tool calls

run-01

    Run in sentences

    run-03

    Quarterly vendor review for Brightmoor Supply

    Not started
    • Thinking0.0s
    • Tools0.0s
    • Waiting on you0.0s
    • Writing0.0s

      The steps show here as the agent takes them

      Trace waterfall

      run-02

      Quarterly vendor review for Brightmoor Supply

      0 calls

      Not started
      Call
      0500ms1s1.5s2s

        The trace fills in as the run makes its first call

        Drag across the time axis to zoom

        Context window bar

        run-05

        Context

        Plenty of room

        80.3k of 200k

        • System6.2k
        • Tools8.6k
        • Conversation12.6k
        • Free103.7k
        • Held for the answer16k

        Context: Plenty of room

        Context cells

        run-06

        Context

        Plenty of room

        80.3k of 200k

        Tokens added per turn2 turns
        Tokens added per turn
        Turn 15.2k
        Turn 27.4k

        Context: Plenty of room

        Agent terminal

        run-04

        Coder~/work/brightmoor-review

        Commands show here as the agent runs them

        Context ring

        run-07

        Check their pricing page as well

        Lumen Swift

        Each message adds its turn to the context

        Context treemap

        run-08

        Context

        Plenty of room

        80.3k of 200k

        Select a tile to see what it holds

        Context: Plenty of room

        Usage by model

        run-09

        Usage

        Illustrative

        $415.20of $500

        83% used
        Limit $500
        Group by
        • Lumen Swift23%$96.40
        • Lumen41%$171.20
        • Lumen Max36%$147.60
        Day 18 of 30Resets on the 1st
        Spend per day, this month
        Day 1$18
        Day 2$21
        Day 3$25
        Day 4$20
        Day 5$12
        Day 6$10
        Day 7$23
        Day 8$26
        Day 9$25
        Day 10$28
        Day 11$23
        Day 12$11
        Day 13$10
        Day 14$28
        Day 15$31
        Day 16$32
        Day 17$29
        Day 18$43

        Agents pause when the limit is reached. Raise it to keep them running until the reset.

        Agent health

        run-10

        Researcher health

        ResearcherLumenIllustrative
        719since last week
        • Success rate

          91.2%
          out of range, target 95% or more
        • Median latency

          3.4s
          in range, target 4s or less
        • Cost per run

          $0.042
          in range, target $0.050 or less
        • Escalations

          7
          out of range, target 5 or fewer
        • Tool errors

          4.8%
          out of range, target 3% or less
        Summary

        A short reading of the five signals for this period.

        Raise the timeout of fetch_page to 20s

        Most failed runs this week stop at the 8s limit on pricing pages.