LibreChat deployment

Place the company endpoint in LibreChat's custom endpoint configuration and load the bearer from the server environment. Do not put the reusable token into a frontend bundle or a committed configuration file.

Example librechat.yaml section:

version: 1.2.1
endpoints:
  custom:
    - name: "Company assistant"
      apiKey: "${B2B_API_KEY}"
      baseURL: "https://api.b2b-llm.com/v1"
      models:
        default: ["gpt-5.4"]
        fetch: true
      titleConvo: true
      titleModel: "gpt-5.4"
      modelDisplayLabel: "B2B LLM"

Replace model values with your enabled selections and choose a suitable utility model for title generation. When supported by the installed version, fetch: true reads the authenticated model inventory rather than relying only on the default list.

Restart the LibreChat service after changing its protected environment or configuration. Verify one user-visible conversation and inspect gateway request history. LibreChat user access is a separate layer from the single credential used by its server; do not assume the gateway automatically receives each chat user's identity.

For distinct project accounting or isolation, provision separate application credentials/endpoints according to your deployment design. Review the installed LibreChat configuration schema before applying options from another release.