Jern Cloud docs

Spend, caps, and usage.

Tokens are your money: sessions run on your own provider key and the provider bills you. Jern bills the workspace fee and machine time. Everything is capped, metered by the gateway from the provider's own usage report, and shown per repository.

What is metered

Every model call an attempt makes crosses the Jern gateway, which holds your provider key and records the usage the provider reports: input, cached input, and output tokens, per model. The count is the provider's, not the agent's. Each usage row is priced at the model catalog's list price on the day, so the dashboard can estimate spend in dollars; the estimate is labelled with the price version it used.

Machine time is metered per size from the moment a machine starts to its destruction and priced per hour; see environments and machines. It is the only cost that grows with the length of a session.

The caps

Cap	Where	What it does
Tokens per attempt	Chosen when the session opens, under the workspace's limit	The gateway refuses the call that would cross it; the attempt ends with what it has and the receipt says the cap was reached.
Monthly token allowance	Settings › Controls	A budget for the workspace, reserved per attempt when it starts. It exists to stop a runaway loop, not to ration.
Dollars per attempt and per month	Settings › Controls	The same guards in money, at list prices, so a team can set one number.
Cap per triggered session	Settings › Triggers	What a label, comment, or schedule may spend per attempt; a schedule can lower it for its own runs.
Time per attempt	Settings › Controls, under the plan's ceiling	The runner publishes what exists and marks the attempt timed out.

The usage report

Under Settings › Usage, the workspace's month: allowance, reserved and remaining tokens, gateway-metered tokens, the estimated spend, machine hours per size and their cost, and attempts by outcome.

Spend by repository

Below the tiles, one row per repository for any of the last six months: sessions, attempts, tokens, model spend, machine time, the total, and beside it the part that schedules started, so unattended work has a visible bill. A month's attempts are those created in it, and machine time follows the attempt. The same numbers download as a CSV with a workspace row, one file per month, and any member of the workspace can read them. The report carries no prompts and no code.

On the pull request

Every pull request's receipt check shows the attempt's tokens against its cap and the estimated spend, so the person reviewing the change sees what it cost without opening the dashboard.

Plans

Every workspace has the full product free for 30 days on its own provider key. After that, $19 a month per workspace for the control plane, plus machine time; tokens bill to your provider account throughout. Plans are bought on GitHub Marketplace or agreed by mail; see Settings.