> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://developers.alephant.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.alephant.io/_mcp/server.

# Model Credentials

> Add and maintain the credentials that let OpenCore access model services securely.

Model credentials give an OpenCore deployment a secure way to reach the model services it is configured to use. Credentials establish access; choose which models are available on the Language Models page.

## How credentials are used

Each credential represents the authentication material for a model service. OpenCore uses it when a configured model or capability sends a request to that service, without exposing the credential to people using chat or Agents.

## Add a credential

1. Sign in as an Admin and open **Models & Capabilities**.
2. Open **Model Credentials** and select the service you want to connect.
3. Enter the credential through the protected configuration form, then save it.
4. Continue to Language Models to select catalog models that use the credential.

## Maintain credentials

Review credentials when a service account changes, access is revoked, or a key is rotated. Update the protected value, confirm the affected models are still available, and remove credentials that are no longer used by the deployment.

## Security practices

Give credentials only the access they need. Do not paste keys into Agent instructions, chat messages, source control, or shared documents.

Use separate credentials when different environments or access boundaries require them. Rotate a credential according to your organization's security process and verify model availability after every change.

## Related pages

* [Language Models](/opencore/models/language-models) explains how to choose catalog models after access is configured.
* [Generation & Chat Capabilities](/opencore/models/generation-and-chat) covers capability-specific configuration and verification.