OpenClaw

OpenClaw supports OpenAI- and Anthropic-compatible endpoints. Prepare a key/model using the shared guide.

1. Install and initialize

Follow Getting started. The reviewed version requires Node.js 24.16+ or 26.1+. Run openclaw onboard --classic and choose custom provider configuration. Use your gateway API key.

2. Add the provider

Open ~/.openclaw/openclaw.json. Merge the following into existing configuration, preserving your channels and agents:

{  "models": {    "mode": "merge",    "providers": {      "vnaipro": {        "baseUrl": "https://api.vnaipro.app/v1",        "apiKey": "YOUR_API_KEY",        "api": "openai-completions",        "models": [          {            "id": "YOUR_MODEL_ID",            "name": "VNAI Pro model"          }        ]      }    }  },  "agents": {    "defaults": {      "model": {        "primary": "vnaipro/YOUR_MODEL_ID"      }    }  }}

Replace the key and both occurrences of YOUR_MODEL_ID. The provider ID is vnaipro; its model reference is vnaipro/YOUR_MODEL_ID.

This example uses Chat Completions. If your model/group requires another protocol:

Use separate provider IDs when configuring multiple protocols. Only declare context/output limits or image support when you know the model's actual capabilities.

3. Apply and test

openclaw doctor
openclaw models list
openclaw gateway restart
openclaw dashboard

If the gateway service is not installed, complete Install the Gateway service in the official guide first. Start a new dashboard chat, verify the model, send a short prompt, and test a file read.

Missing model: check that the provider's model list and model reference match. Old key still used: check the active authentication profile/provider and restart the gateway. See common errors.

Sources: Custom providers, Models.