Skip to main content
Continue supports per-model request headers in config.yaml — add a Toolken model entry, set requestOptions.headers, and every Continue request is tracked with full cost and latency visibility.
1

Get your two keys

You need two keys:Toolken forwards your provider key verbatim and never stores it.
2

Add the Toolken model to config.yaml

Open Continue’s config.yaml (usually ~/.continue/config.yaml) and add a model entry under models:
apiKey is your provider key (BYOK). It is sent as Authorization: Bearer and forwarded untouched. X-Toolken-Key authenticates you to Toolken and is stripped before the request reaches your provider.
Set X-Toolken-Metadata-Agent to any string that identifies this model in your dashboard. You can define multiple Toolken model entries with different X-Toolken-Metadata-Agent values to break spend down by use case.
3

Select the model in Continue

Open the Continue panel in your editor, switch to the Toolken GPT-4o mini model (or whatever you named it), and make a request. Continue attaches X-Toolken-Key and X-Toolken-Metadata-Agent automatically via requestOptions.headers.
4

Confirm in your dashboard

Within seconds your request appears in the Toolken dashboard, grouped under the continue agent. Cost, token usage, and latency are all captured automatically.

​
Next

Codex

Route Codex CLI traffic through Toolken with a custom provider entry.

Providers & Routing

One gateway URL, 13 providers, your own keys.