Skip to main content
Despite the name, the API Keys tab (/enterprise/orgs/{org_id}/api-keys) doesn’t manage generic org API keys — it’s the configuration screen for the Enterprise LLM Gateway: storing provider credentials, discovering provider models, and issuing team-scoped gateway keys.
This tab configures the hosted gateway at https://proxy.reclaimllm.com/<org-slug>/v1. It is not related to the developer Local API Proxy (rclm-proxy on http://localhost:9800).

​
What you’ll see

  • Provider Credentials cards for OpenAI, Azure OpenAI, Anthropic, Gemini, Moonshot, and DeepSeek, each showing whether the provider is configured or disabled.
  • Models under each provider, collapsed by default, showing the provider model names available to that org.
  • Sync models for configured providers, so admins can refresh the org’s model list after a provider releases new models.
  • Team Gateway Keys — a creation form (name, team, allowed providers, and optional concrete model allowlists) and a table of existing keys (Name, Team, Providers, Last used, Revoke/Delete).
  • A copy-pasteable curl example for calling the gateway (OpenAI, Moonshot, Azure OpenAI chat, and Azure embeddings).

​
Proxy request examples (curl)

These match the Use this proxy with curl panel on the LLM Gateway page. Replace acme with your org slug and use your team gateway key.

​
OpenAI chat

​
Moonshot chat

​
Azure OpenAI chat

​
Embeddings (Azure OpenAI)

Request models must be LiteLLM-prefixed. Org catalog and key allowlists store the bare name. See the model naming contract.

​
Provider credentials and model discovery

When an admin saves a provider API key, ReclaimLLM calls that provider’s list-models API from the backend using the saved credential. Any returned model names that are not already present for the org are inserted into the org’s gateway model registry. The stored model name is the provider’s returned model name as-is (bare, no LiteLLM prefix). Clients must still call the gateway with a prefixed request model (for example catalog gpt-5.6-luna → request openai/gpt-5.6-luna). See the model naming contract.

​
Syncing models later

Use Sync models on a configured provider to refresh the model registry at any time. The sync only adds new model names for that provider and org; existing model rows are left unchanged. The sync result is collapsed by default. Expand it to inspect the model names returned by the provider API.

​
Team gateway keys and model access

Team gateway keys can be scoped to providers and, when needed, specific bare model names from the org’s discovered model list. Leaving a provider’s model selection empty allows all enabled models for that provider; selecting models restricts that gateway key to only those catalog names.

​
What you can do

Save, replace, enable, disable, or remove provider credentials; sync provider models; enable, disable, add, or delete org model rows; create a team-scoped gateway key; revoke or delete a key.

​
Who sees it

Admin only. Full setup steps, the model naming contract, and request examples are covered in the Enterprise LLM Gateway guide.