Usage, limits and costs

Gloop's agents are Claude Code, and they use your Claude plan as in any terminal. Gloop adds no costs of its own: it shows you how much you're using, because with several agents at once the limits come sooner.

#The bar at the bottom of the column

At the bottom of the column, Usage shows the fullest window of the plan (or the one you choose in Settings → Agents): the 5 hours or the week, with the percentage and when it resets. The colour grows with it: green up to 40%, then yellow, red and burgundy at 100%. A click opens the Usage window.

#The Usage window

• Your plan: the 5-hour session and the week, each with the percentage used and when it resets.
• Today's cost: what today's sessions would cost at API prices, the way Claude Code counts it.
• Agents' context: for every open agent, the workspace, the model, what its session costs and how much context it has used.

#The cost, with a subscription

The cost is at API prices: with a Pro or Max subscription you don't pay it, except for the part beyond the plan's limits with extra credits on. It's there to understand how much a piece of work weighs, and to compare models and ways of working (see The right model for the task and a Scout request's cost).

When a window of the plan is full:

• with extra credits on (/extra-usage in a pane) the agents carry on, and what they spend until the reset is paid at API prices. The Usage window says how much you've spent beyond the limit;
• without, the agents stop until the reset.

If you use Claude through an organisation with a spending limit, Usage shows that: how much you've spent out of how much, per day, week or month.

#When you reach a limit

An agent stopped by the limit ends its turn, and the pane shows the reset time and the Resume on its own switch: when on, a minute after the reset Gloop tells the agent to carry on (the app must be open; a message of yours before then cancels it). See Agents.

The coordinator doesn't open new agents when a window of the plan is over 85%.

#Context

The context is the memory of an agent's conversation. The percentage used shows at the bottom of every pane and in the Usage window. Above 70% it's worth wrapping the conversation up: a full context makes Claude slower and less precise. Gloop offers to continue in a new session with a handoff: see When the context fills up.