LibreChat

LibreChat supports custom OpenAI-compatible endpoints. The common approach is to define the endpoint in librechat.yaml and store the key in .env.

Preparation

Field	Value
API URL / Base URL	https://n.tokeness.dev/v1
API Key	A key created in the Tokeness console
Models	Full model names copied from the Tokeness model catalog

LibreChat is typically used for multi-user chat. In production, do not expose the Tokeness key to the frontend and do not commit it to a repository.

File Responsibilities

A LibreChat custom endpoint typically involves three files:

File	Purpose
librechat.yaml	Defines the custom endpoint, model list, and display name
.env	Stores sensitive information such as API keys
docker-compose.override.yml	Mounts librechat.yaml in Docker deployments

If you are not using Docker, place librechat.yaml wherever your deployment can read it.

Configure .env

Add the following to .env:

TOKENESS_API_KEY=your Tokeness API key

Do not put the real key in librechat.yaml.

Configure librechat.yaml

Below is a minimal configuration. Replace the model names with the full model names from the Tokeness model catalog.

version: 1.2.1

endpoints:
  custom:
    - name: "Tokeness"
      apiKey: "${TOKENESS_API_KEY}"
      baseURL: "https://n.tokeness.dev/v1"
      models:
        default:
          - "YOUR_MODEL_NAME"
        fetch: false
      titleConvo: true
      titleModel: "YOUR_MODEL_NAME"
      modelDisplayLabel: "Tokeness"

Field descriptions:

Field	Description
name	The endpoint name displayed in LibreChat
apiKey	Reads the Tokeness key from .env
baseURL	The Tokeness OpenAI-compatible Base URL
models.default	Models available in the UI
models.fetch	When set to false, uses the manual model list
titleModel	The model used to generate conversation titles

To expose multiple models:

models:
  default:
    - "MODEL_A"
    - "MODEL_B"
    - "MODEL_C"
  fetch: false

Docker Mount Configuration

If you use Docker Compose, you can mount the configuration in docker-compose.override.yml:

services:
  api:
    volumes:
      - ./librechat.yaml:/app/librechat.yaml

After making changes, restart:

docker compose down
docker compose up -d

Let Users Provide Their Own Key

For an internal team deployment where each user should use their own Tokeness key, change apiKey to:

apiKey: "user_provided"

In this mode, users enter their own key in the LibreChat UI. It suits scenarios where multiple people share a LibreChat instance but you do not want the server to hold a shared key.

Verification

1. Restart LibreChat.
2. Select Tokeness in the model endpoint selector.
3. Select a configured model.
4. Send "reply with ok only".
5. Open the Tokeness usage logs and confirm a request arrived.

Troubleshooting

Symptom	Action
Tokeness endpoint not visible	Check whether librechat.yaml was mounted successfully and restart the container
401	Check that TOKENESS_API_KEY in .env is being read by the container
404	baseURL should be https://n.tokeness.dev/v1
Model dropdown is empty	Set models.fetch: false and enter model names manually
Conversation title generation fails	Check that titleModel is an available model
Multiple users' keys get mixed up	Use user_provided so each user enters their own key in the UI

References

• LibreChat Custom Endpoints
• LibreChat Custom Endpoint Object Structure