Using NVIDIA's OpenAI-Compatible API in OpenClaw

NVIDIA provides free access to open models through an OpenAI-compatible API at https://integrate.api.nvidia.com/v1, authenticated with an API key from build.nvidia.com. OpenClaw sets the NVIDIA provider's default to Nemotron 3 Ultra, NVIDIA's 550B total / 55B active reasoning model built for long-context agentic tasks.

Getting started

Get your API key

Generate an API key at build.nvidia.com.

Export the key and run onboarding

export NVIDIA_API_KEY="nvapi-..."
openclaw onboard --auth-choice nvidia-api-key

Set an NVIDIA model

openclaw models set nvidia/nvidia/nemotron-3-ultra-550b-a55b

For non-interactive configuration, pass the key directly:

openclaw onboard --auth-choice nvidia-api-key --nvidia-api-key "nvapi-..."

Warning

--nvidia-api-key puts the key in shell history and ps output. Use the NVIDIA_API_KEY environment variable whenever possible.

Config example

{
  env: { NVIDIA_API_KEY: "nvapi-..." },
  models: {
    providers: {
      nvidia: {
        baseUrl: "https://integrate.api.nvidia.com/v1",
        api: "openai-completions",
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "nvidia/nvidia/nemotron-3-ultra-550b-a55b" },
    },
  },
}

Featured catalog

When an NVIDIA API key is configured, setup and model-selection paths retrieve NVIDIA's public featured-model catalog from https://assets.ngc.nvidia.com/products/api-catalog/featured-models.json and cache the result for 24 hours (first 32 entries, imported as free text-input rows). New featured models from build.nvidia.com then show up in setup and model-selection surfaces without needing an OpenClaw release. If the live feed is available, the first returned model becomes the preselected option during NVIDIA setup.

The fetch enforces a fixed HTTPS host policy for assets.ngc.nvidia.com. If no NVIDIA API key is configured, or if the feed is unavailable or malformed, OpenClaw falls back to the bundled catalog and the bundled default below.

Nemotron 3 Ultra

Nemotron 3 Ultra is OpenClaw's default NVIDIA model. NVIDIA's build page for nvidia/nemotron-3-ultra-550b-a55b lists it as a free endpoint with a 1M-token context specification.

The bundled Ultra row sends chat_template_kwargs: { enable_thinking: false, force_nonempty_content: true } by default so normal chat output stays in the visible answer rather than exposing reasoning text.

Use Ultra for the highest-capability NVIDIA default. Keep Super selected when you want the smaller Nemotron 3 option, or pick one of the third-party models hosted in NVIDIA's catalog when their context, latency, or behavior suits your needs better.

Bundled fallback catalog

The selectable bundled rows are a snapshot of NVIDIA's featured-model catalog. Deprecated compatibility rows remain resolvable by exact reference but do not appear in model pickers.

The full compatibility catalog also keeps these shipped refs for existing configurations: nvidia/moonshotai/kimi-k2.5, nvidia/z-ai/glm-5.1, nvidia/minimaxai/minimax-m2.5, nvidia/z-ai/glm5, and nvidia/minimaxai/minimax-m2.7. They remain available by exact reference but never show up in onboarding or model pickers.

Advanced configuration

Auto-enable behavior

The provider activates automatically when the NVIDIA_API_KEY environment variable is set or a key was stored during onboarding. No explicit provider configuration is needed beyond the key.

Catalog and pricing

OpenClaw prefers NVIDIA's public featured-model catalog when NVIDIA auth is configured and caches it for 24 hours. The bundled selectable fallback is a static snapshot of NVIDIA's featured-model catalog; deprecated exact-reference compatibility rows are hidden from model pickers. Costs default to 0 in source since NVIDIA currently provides free API access for the listed models.

OpenAI-compatible endpoint

OpenClaw communicates with NVIDIA using the openai-completions adapter against the standard /v1 chat completions route. Any OpenAI-compatible tooling should work out of the box with the NVIDIA base URL.

Nemotron 3 Ultra reasoning params

NVIDIA's Ultra sample request uses chat_template_kwargs.enable_thinking and reasoning_budget for reasoning output. OpenClaw's bundled Ultra row disables template thinking by default for normal chat use. If you need to opt into NVIDIA reasoning output or force other NVIDIA-specific request fields, set per-model params and keep provider-specific overrides scoped to the NVIDIA model:

{
  agents: {
    defaults: {
      models: {
        "nvidia/nvidia/nemotron-3-ultra-550b-a55b": {
          params: {
            chat_template_kwargs: { enable_thinking: true },
            extra_body: { reasoning_budget: 16384 },
          },
        },
      },
    },
  },
}

params.chat_template_kwargs merges into any chat_template_kwargs already on the request instead of replacing the entire object. params.extra_body is the final OpenAI-compatible request-body override and overwrites colliding payload keys, so use it only for fields NVIDIA documents for the selected endpoint.

Slow custom provider responses

Some NVIDIA-hosted custom models can take longer than the default ~120s model idle watchdog before they emit a first response chunk. For custom NVIDIA provider entries, raise the provider timeout instead of the whole agent runtime timeout; timeoutSeconds covers provider HTTP requests and raises the idle/stream watchdog ceiling for that provider:

{
  models: {
    providers: {
      "custom-integrate-api-nvidia-com": {
        baseUrl: "https://integrate.api.nvidia.com/v1",
        api: "openai-completions",
        apiKey: "NVIDIA_API_KEY",
        timeoutSeconds: 300,
      },
    },
  },
  agents: {
    defaults: {
      models: {
        "custom-integrate-api-nvidia-com/meta/llama-3.1-70b-instruct": {
          params: { thinking: "off" },
        },
      },
    },
  },
}

Tip

NVIDIA models are currently free to use. Check build.nvidia.com for the latest availability and rate-limit details.

Related

• Model selection, Choosing providers, model refs, and failover behavior.

• Configuration reference, Full config reference for agents, models, and providers.