Coding tools

OpenClaw

Use RuToken as the model provider for the self-hosted personal AI assistant OpenClaw.

OpenClaw is an open-source, self-hosted personal AI assistant. A single Gateway process connects messaging apps such as Telegram, Discord, WhatsApp and iMessage to AI agents running on your own hardware, with long-term memory, scheduled tasks, tool calling and multi-agent routing. Docs: docs.openclaw.ai, source: GitHub.

Prerequisites

An API key: open the console, go to API Keys and create a dedicated key (for example, named OpenClaw).
Node.js 22 or newer.

Installation

Install OpenClaw and run the onboarding wizard (for other methods see Getting Started):

bash

curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard --install-daemon

Check that the Gateway and the Control UI work before adding a provider — this makes later troubleshooting much easier:

bash

openclaw gateway status
openclaw dashboard

Configuration

OpenClaw adds OpenAI-compatible gateways through models.providers in its config file, usually ~/.openclaw/openclaw.json (the location can be changed with OPENCLAW_HOME, OPENCLAW_STATE_DIR or OPENCLAW_CONFIG_PATH).

1. Provide the API key

Store the key in an environment variable that the Gateway process can read (current shell, service environment, or an .env file read by OpenClaw):

bash

export NEWAPI_API_KEY="sk-xxxx"

2. Add the provider

Merge this snippet into openclaw.json and list the model IDs you want to use:

json5

{
  models: {
    mode: "merge",
    providers: {
      newapi: {
        baseUrl: "https://rutoken.ai/v1",
        apiKey: "${NEWAPI_API_KEY}",
        api: "openai-completions",
        models: [
          { id: "gpt-4o-mini", name: "GPT-4o mini" },
          { id: "gemini-2.5-flash", name: "Gemini 2.5 Flash" },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: {
        primary: "newapi/gpt-4o-mini",
        fallbacks: ["newapi/gemini-2.5-flash"],
      },
    },
  },
}

Key	Description
`models.mode`	`merge` keeps OpenClaw's built-in providers and adds this one
`providers.newapi.baseUrl`	`https://rutoken.ai/v1` — the `/v1` suffix is required
`providers.newapi.apiKey`	the key, preferably injected as `${NEWAPI_API_KEY}`
`providers.newapi.api`	`openai-completions` for OpenAI-compatible gateways
`providers.newapi.models`	model IDs exactly as shown in the Model Square
`agents.defaults.model.primary`	default model in `provider/model-id` format
`agents.defaults.model.fallbacks`	models to try if the primary one fails

Verify

Restart the Gateway, then check that the new models are listed and chat in the Control UI:

bash

openclaw models list
openclaw dashboard

Models with the newapi/ prefix should appear, and the default model should be newapi/.... Requests show up in the console under Usage Logs.

Troubleshooting

Symptom	Fix
`401` / invalid API key	Make sure the Gateway process — not only your terminal — can read `NEWAPI_API_KEY`, especially when it runs as a background service.
`404` / not found	`baseUrl` must end with `/v1`.
Model not found	`primary` and `fallbacks` must reference an `id` listed under `providers.newapi.models`, and that ID must exist in the Model Square.
Unclear errors	Run the Gateway in the foreground to see logs: `openclaw gateway --port 18789`.

← Previous

CC Switch

Cherry Studio