Usage and logs

Every model call through the AI Gateway is one request with one record. Usage shows spend and Logs shows the requests, both under Platform > AI governance. Each person sees their own spend in Settings > Usage.

​

Usage

Pick a range, then Group by person, group, API key, model, client, provider or agent, and on xpander cloud and Hybrid also license model. View requests on a row opens Logs with the same filters, and Export downloads the table as CSV. Amounts switch between Credits and $. The top row shows Model cost: what the calls were worth at xpander’s model prices, not the credits billed. When more than one payer has calls in the range, one tile per payer follows: Xpander AI Credits (xpander’s key), Your provider keys (with the kinds of credential you stored, such as API key or AWS role) and Subscription: personal Claude plans. Costs are estimated from token counts and xpander’s model prices, including cache reads and writes. For calls on your own key, your provider’s invoice is exact.

​

Logs

Logs lists requests newest first, grouped into traces when they belong together, such as one agent run or one coding session. Live streams new requests. Each row shows the Person, the Source (what started it, then the coding tool or harness it ran on), the agent, model, tokens, cost, Paid by (for example Your provider keys, Xpander AI Credits or a person’s Claude plan), duration and status. A refused call shows the reason, such as Budget reached, Model not allowed or No credential. Filter narrows by person, model, status, group, API key, agent, what started the run (Started by), client, Paid by (a license model, or one of your stored provider keys) or xpander operation. Filters stay in the page address, so a filtered view can be shared as a link. To open one request, search its request or trace ID. Each request has its own link: https://platform.xpander.ai/governance/logs/<request or trace ID>. Export downloads the filtered list as CSV, JSON or a table to paste. Exports never include prompt or response text.

​

A trace or a request

A trace shows who, the agent and channel, the time range, turns, calls, credits and what paid, then each turn with its calls. A running trace shows Live. A request shows its model, latency, tokens and estimated cost, plus:

• Routing: the Requested model, the Alias it matched, what it Resolves to, Fallbacks and each of the Attempts.
• Gateway decisions: every check the call met (Own plan, Credits, Budget, Limits, Request rate, Token rate, Parallel calls) and its verdict.
• Prompt and response, when text is saved: the conversation and the raw request and answer.

Text is saved for everyone unless an admin turns off Saved prompt text. Roles that may view usage can open it, and every opened request is recorded in the Audit log. Log destinations never include it.

​

Requests xpander makes for you

Some features call a model on your organization’s behalf: conversation titles and summaries, building an API connector, generating an agent’s details, checking custom function code, image generation and transcription in built-in skills, and bot replies. They appear in Logs like any other request, under the person who started the work (or -), with the operation’s step as the source. Each shows its cost and counts under the key that ran it. On xpander’s own key it adds to Xpander AI Credits in Usage, but is never charged in credits and counts toward no budget. On your own provider key your provider bills it, which is always the case on a self-hosted install. Operations charged in credits, such as scoring answers in evaluations, are billed as usual. To list only these requests, choose System calls.

​

How long requests are kept

Set by Keep requests for in Policies. An older request opens as No request or trace with this ID. Requests live in your organization’s own database: a dedicated database on xpander cloud, or your install’s PostgreSQL when self-hosted. On Hybrid, requests from agents on your install stay in your environment. If your browser can’t reach it directly, Logs shows Needs a direct connection to your environment. Usage totals and the calls people’s coding tools make through xpander cloud always show.

​

Next steps