Usage and Billing
For AI agents: see llms.txt for the complete documentation index. Markdown versions are available by adding .md to a page URL or requesting Accept: text/markdown.
The gateway records the project and function behind every request automatically. You don't need to set anything up.
Seeing your usage
The team usage page has an AI Gateway section that shows daily spend per project. Open it from the AI Gateway card on the usage summary.
To see which functions spent the money, scroll to the breakdown by function and pick the AI tab. It lists spend per function.
See the usage page docs for the rest of the page.
Usage limits
Set daily or monthly AI Gateway limits for a cloud deployment from Settings → Usage Limits. Choose the AI Gateway metric and set a disable threshold in dollars. Deployments other than development deployments also support warning thresholds. Reaching the disable threshold disables the whole deployment, not just AI Gateway requests.
See Usage limits for thresholds and reset windows.
How billing works
We charge the same rates as OpenRouter. Price per token depends on the model, when you use it, and whether the tokens are cached. Image and video prices depend on the model and generation options, such as resolution and duration. Voice models charge per character of input, per second of audio, or per token. Cancelling a video request after submission can still incur its generation cost. The charge appears as an AI Gateway line item on your Convex invoice, in dollars.
Spending limits apply to AI Gateway spend like any other usage. Set a limit on the team billing page to cap what your team pays per month.
Requests from a project-connected local deployment are attributed to that project's team and count toward the same team spending limit. Local deployments do not have a separate deployment-level live usage meter. Spending-limit state is checked when the gateway credential is minted, so an in-flight request or an already issued short-lived credential may finish after the limit is reached.