AnvilHub

Your crew & fleet

Runs — the record of work

Every piece of work your crew does is recorded — what ran, what it changed, what it cost. The terminal shows it once; AnvilHub keeps it.

When an agent does something in the terminal, you watch it happen and then it scrolls away. The Runs page on AnvilHub is the opposite: a durable, searchable record of every run across your whole fleet, kept long after the session ended.

What a run records

Each run on the page carries:

  • Who ran it — which bot or person, and from where (a session, a routine, a chat).
  • What happened — its state and how long it took.
  • What it changed — the changeset, and any artifacts it produced.
  • What it needed — any approvals along the way.
  • What it cost — tokens and spend, with anything the provider didn't price marked honestly as unpriced rather than silently counted as zero.

Grouping and export

The record is only useful if you can slice it. Runs can be grouped by bot, by day, or by routine, so "what has the release bot been doing this week" and "what did everything cost yesterday" are both one view away. Filters narrow it further, there's an audit drawer for the privileged actions, and you can export to CSV when you need the numbers somewhere else.

Cost you can trust

Runs is also where cost stops being a per-session surprise. Because every run reports its tokens and spend, the page can roll the total up across bots, days, routines and people — across every connected machine. The Spend settings pane does the same per instance. The honesty rule matters here: an unpriced model's usage is shown as unpriced, never folded into a falsely precise total.

In the terminal

The same underlying record is reachable with /history, /runs, and /audit. The difference is persistence and perspective: the terminal shows you one machine's recent activity; AnvilHub keeps the whole fleet's history and lets you group and total it.

Where to next