Your agents run on your own subscriptions, and each subscription has usage windows. The usage policy says how much of a window work on a project may use and which client takes the next iteration of a loop. Overlane does not start an agent itself: the loop you run on your own machine (overlane run loop) and the agents you connect read the policy and stop when it says to.
Open it under Settings, Agents, Usage policy. It shows the cap that applies, each client and whether it can take the next iteration, the project’s policy and your own settings.
The cap that applies
The cap is the lowest of three numbers: the project cap, your own cap, and 100% minus the share you keep free for your own work. Each client’s row says whether it can take the next iteration now and, if not, why.
Project policy
| Setting | Choices | What it does |
|---|---|---|
| Clients | One after another, Take turns, At the same time | The order clients take iterations. At the same time also has a limit on how many run at once |
| Pace | Use it up to the cap, Spread across the window | Use usage as fast as allowed, or spread it across the window |
| Cap | 1 to 100 | The most of each usage window that work for this project may use |
Your settings
| Setting | What it does |
|---|---|
| Your cap | The most of each usage window to use on your subscriptions |
| Kept free for you | The share of each window left for your own work, 0 to 99 |
| Share usage events | Off unless you turn it on. An event holds the client, how the iteration ended and the usage it took, with no prompt, code or output |
From the CLI
overlane usage policy show
overlane usage policy set --cap 80 --pace steady
overlane usage policy set --scope user --reserve 20 --share offCommand reference: usage policy show and usage policy set. The loop that follows the policy is described in run loop.