Watching your agents
See the work, the cost, and the behavior of every run.
Running agents you can't see is a leap of faith. Fragua is built so you never take it: every run is watchable along three axes — the work it's doing, what it's costing, and how it behaved. This is the deepest-built part of the app, and the reason a team can run many agents without losing the thread.
A live log of what the agent is doing, turn by turn, as it happens.
Tokens and dollars for every turn, rolled up across runs, workspaces, and time.
A durable record of what each run touched, ran, and why it stopped.
Watch the work as it happens
Open a run and you see a live log that streams, not polls — the server pushes each new line over the connection as the agent produces it, so the page fills in on its own. You see the agent's messages, the questions it asks mid-interview, the tools it runs and what they returned, and the draft when it lands. Every run gets the full firehose; the page decides what to show, so the conversation stays readable while the complete detail is one click away on the run's own page.
The log is durable. A session is pinned to each run from the start, so the work survives a closed browser or a crash — come back tomorrow and the log is still complete, right where it left off.
Every token, accounted for
Fragua reads the usage from every turn into its own record — the model, the token counts, the cost — and rolls it up so you always know where the money went: per run, per workspace, per account, per month, and per phase across research, plan, foundation, spec, and execution.
This is metering, not gating — there are no spend caps to trip over. And it's your bill, not ours: agents run on your own AI key on your machine, so tokens bill straight to your provider. Fragua never sits in the path of the call, so we can't see your bill or charge you for usage.
The record of how each run behaved
The same log doubles as an audit trail. Every run carries the same fields — what it touched, what it ran on, why it ended — so a stalled or failed run is never a mystery. A heartbeat watchdog notices a run that's gone quiet and surfaces it, and runs you discard or restart still count toward the cost roll-ups, so a budget can't be reset by starting over.
You don't have to sit and watch
Fragua tells you when something needs you. Notifications land in the in-app tray, pop a toast while you're working, and — if you've turned it on — reach you as a web push even when the tab is closed.
-
A draft is ready — when a run finishes and there's something to review.
-
A run needs you — when the agent has questions and is waiting on your answer.
-
A run failed — when something went wrong, with the reason and a link to the log.
See exactly what's kept, and where your machines connect.