OpenClaw

OpenClaw is an open-source personal AI assistant gateway: one process that runs your agents, connects them to chat channels, and manages tools and sessions. It ships with a provider catalog covering the big vendors, and any OpenAI-compatible endpoint can be registered as a custom provider. Register Nebul, and your agents run on your project's models.

Prerequisites

Node.js 20 or later
A Nebul AI Studio account with API access
An API key for your active project
A model ID from the Model Catalog

Installation

npm install -g openclaw@latest --allow-scripts=openclaw

On npm 11.15 and earlier, omit --allow-scripts=openclaw. Then run onboarding:

openclaw onboard --install-daemon

The wizard verifies model access, creates the workspace, and configures the gateway. See the getting started guide for the full walkthrough, including chat channels like Telegram and Discord.

Connect it to Nebul

Custom providers live under models.providers in your OpenClaw config, or in ~/.openclaw/agents/<agentId>/agent/models.json. Register Nebul like this:

{
  models: {
    mode: "merge",
    providers: {
      nebul: {
        baseUrl: "https://api.inference.nebul.io/v1",
        apiKey: "sk-your-api-key-here",
        api: "openai-completions",
        models: [
          {
            id: "zai-org/GLM-5.3",
            name: "GLM 5.3",
            reasoning: true,
            input: ["text"],
            contextWindow: 1048576,
            maxTokens: 16384
          }
        ]
      }
    }
  }
}

Field notes:

api selects the request format. openai-completions is the Chat Completions format; openai-responses also works with Nebul if you prefer the Responses API.
The id must exactly match a catalog model ID. Model references elsewhere in OpenClaw then read nebul/zai-org/GLM-5.3.
contextWindow and maxTokens shape context management. Take contextWindow from /v1/model/info; a conservative output cap like the one above is fine.
reasoning: true enables the thinking controls for models that support them.

Prefer not to edit files by hand? The same block can be set from the command line:

openclaw config set models.providers.nebul '<json>' --strict-json --merge

For worked examples of custom providers, see the provider examples in the OpenClaw docs.

Run it

After a config change, restart the gateway or run the agent again. Pick the nebul/zai-org/GLM-5.3 model in the dashboard or wherever your agent's model is selected, then send a message. Check the route first if you like:

openclaw gateway status

Troubleshooting

404 on requests: baseUrl must be https://api.inference.nebul.io/v1, with the /v1 suffix.
Model not selectable: model IDs in OpenClaw are provider_id/model_id. Confirm the id matches a catalog ID exactly.
Requests rejected with unusual fields: Nebul accepts the developer role, max_completion_tokens, and reasoning_effort, so the defaults work. If you copied compat flags from another provider's example, remove them; OpenClaw's own catalog owns those for known routes.

Data and telemetry

OpenClaw's anonymous feature statistics are off by default, but it does send a daily update check that includes your OpenClaw version, OS, Node version, and CPU architecture. Two controls, from least to most thorough:

openclaw telemetry off

That disables the optional statistics. To stop the daily update check as well:

{
  update: {
    checkOnStart: false
  }
}

Run openclaw telemetry show to see exactly what would be sent. See the OpenClaw telemetry page for the full breakdown.

Model traffic goes only to the providers you configure. For how Nebul treats requests, see Privacy & Security.