Every Valar token, costed.

Create an API key in the Valar dashboard at app.valarhq.ai. Then open Data ingestion and integrations > Integrations, find Valar, enter the key, and test the connection before saving. Either an Inference key or a Coding key works, because both read the same usage data.

An Inference key or a Coding key both work. We recommend a dedicated key with an expiry that you use only for this connection.

Yes. ValarCode traffic is billed on its Coding keys and comes through the same usage data as the rest of your inference, so it appears once, under those keys.

We refresh your Valar usage daily. Valar reports usage by UTC day, so we import each day once it closes.

Charges map to Provider, Project (the API key ID), Service (Model Inference) and SKU (the model that served the request, which can differ from the model you asked for when a router picks it). Labels include model, model provider, API key ID and API key name, and every row carries input, cached and output token counts. Workload tags are not imported, because Valar's usage API does not return them.

Valar reports spend in whole cents at every level, so per-model rows can add up to slightly less or more than the dashboard total because of rounding. Taxes, credits and refunds are not in Valar's usage API, so Cloud Analytics shows your usage spend.

No. Requests that ValarCode serves on your own provider credentials are billed by that provider, not by Valar, so they are not in your Valar spend. Connect that provider to see them.

Connect once per Valar organization. The usage API returns spend for the whole organization, so connecting a second key from the same organization would count the same costs twice.