Skip to main content
If you already govern AI traffic with LiteLLM Proxy or Kong AI Gateway, Governance → Switch imports that configuration into Caveman Cloud so you do not rebuild keys, budgets, and teams by hand. Importing requires the Owner, Platform admin, or Developer role.

​
Supported sources

​
What gets created

​
Run an import

1

Choose the source format

Open Governance → Switch and pick LiteLLM Proxy or Kong AI Gateway.
2

Paste the configuration

Paste the YAML. For LiteLLM, you can also enter the server URL and master key to read keys, teams, users, budgets, and guardrails. Caveman only reads from the server; it never changes it.
3

Review the plan

The What Caveman will create panel lists every provider, model, limit, project, key, and member before anything is written.
4

Add provider secrets

Config files reference provider keys through environment variables such as os.environ/OPENAI_API_KEY. Paste the secret for each variable so Caveman can store it as a provider connection.
5

Apply and copy new keys

Apply the import. New key secrets are shown once: copy them and update your apps.
Key secrets cannot be copied over because the source system stores them hashed. Every imported key gets a new secret, so plan a cutover window to swap credentials in your apps.
Re-running an import is idempotent, so you can safely re-apply a partial run.

​
After the import

  • Check Governance → Budgets and Governance → Keys to confirm caps and per-key limits match what you expect.
  • Add team and cost-center tags to imported keys for cost allocation.
  • Point your apps at the Caveman gateway URL with their new keys. See Authentication.

​
Next steps