Model access and provider connections

A model route pairs a model with the provider that will serve it. The same model may be available through more than one route. Choose a route based on your organization's available connections and how you want usage billed.

Connect a provider

An administrator opens provider settings, selects the provider, enters the credential in the dashboard, and completes the connection. Inspect its status before selecting it in a prompt. When replacing a key, update the provider connection and confirm its status before retiring the old credential at the provider.

CLI, API, and MCP provider-setup operations return a browser handoff. Open that PromptLens URL to enter secrets in the dashboard. Keep credentials out of shared command output and assistant transcripts.

Screenshot placeholder — Provider and model selection Show a connected provider and the corresponding available model route; conceal credential values.

Choose how calls are funded

An organization provider key takes precedence for its route. Those calls are billed by the provider and do not spend PromptLens Credits. Eligible OpenRouter routes can use PromptLens Credits when no overriding organization connection is used and sufficient balance is available.

A catalog listing describes a route; it does not prove that your organization currently has usable access or credit for it. Check the selected route and connection when a run reports unavailable model access.

Configure generation and judging separately

The prompt's generation model produces the answer. Each LLM score or checklist evaluator selects its own judge model and generation settings. Provide usable access for both. Changing one does not automatically reconfigure the other.

Removing a connection can affect prompts and judges that rely on it. Review those dependencies before removal, then select another usable route or funding option for future runs. See credits for reservations and usage charges.