Skip to main content
LiteLLM support is flag-gated. Ask your Cloud Capital Customer Success Manager to enable it for your organization.
Cloud Capital reads daily activity from your LiteLLM proxy and imports spend by team or user, upstream provider, model, and key alias. LiteLLM spend is modelled at list price unless your proxy defines custom pricing. The key is used only to read cost activity.
Connect either LiteLLM or the upstream OpenAI or Anthropic billing integration for the same traffic. Cloud Capital preserves every source and does not deduplicate overlapping rows.

​
Requirements

  • A publicly reachable HTTPS LiteLLM base URL on port 443. Private, loopback, link-local, local DNS names, redirects, and other non-global destinations are refused.
  • A proxy_admin_viewer key restricted from inference. Configure models: ["no-default-models"] or max_budget: 0.

​
Setup

1

Create a restricted viewer key

Create a LiteLLM proxy_admin_viewer key and restrict it from inference. Cloud Capital calls only GET /team/daily/activity.
2

Create the integration

Choose LiteLLM, enter the public base URL and viewer key, then select Test Connection. The test makes the same guarded daily-activity read as the nightly loader.
3

Enable and save

After a successful test, set the integration to Enabled and save it. Daily ingestion runs on the shared vendor-billing runtime.
LiteLLM spend does not appear in Cost Insights or forecasts yet. The integration loads and aggregates daily LiteLLM spend.