Running OpenClaw with vLLM: OpenAI-Compatible Local Server

vLLM exposes open-source (and some custom) models through an HTTP API that is OpenAI-compatible. OpenClaw connects using the openai-completions API and can auto-discover models when you opt in with VLLM_API_KEY.

Getting started

Start vLLM with an OpenAI-compatible server

Your base URL must expose /v1 endpoints (/v1/models, /v1/chat/completions). vLLM typically runs on:

http://127.0.0.1:8000/v1

Set the API key environment variable

Any non-empty value works if your server does not enforce auth:

export VLLM_API_KEY="vllm-local"

Select a model

Replace with one of your vLLM model IDs:

{
  agents: {
    defaults: {
      model: { primary: "vllm/your-model-id" },
    },
  },
}

Verify the model is available

openclaw models list --provider vllm

Tip

For non-interactive setup (CI, scripting), pass the base URL, key, and model directly:

openclaw onboard --non-interactive \
  --mode local \
  --auth-choice vllm \
  --custom-base-url "http://127.0.0.1:8000/v1" \
  --custom-api-key "vllm-local" \
  --custom-model-id "your-model-id"

Model discovery (implicit provider)

When VLLM_API_KEY is set (or an auth profile exists) and models.providers.vllm is not defined, OpenClaw queries GET http://127.0.0.1:8000/v1/models and converts the returned IDs into model entries.

Note

If you set models.providers.vllm explicitly, OpenClaw uses only your declared models. Add "vllm/*": {} to agents.defaults.models to make OpenClaw also query that configured provider's /models endpoint and include all advertised vLLM models.

Explicit configuration

Configure explicitly when vLLM runs on a different host or port, you want to pin contextWindow/maxTokens, your server requires a real API key, or you connect to a trusted loopback, LAN, or Tailscale endpoint:

{
  models: {
    providers: {
      vllm: {
        baseUrl: "http://127.0.0.1:8000/v1",
        apiKey: "${VLLM_API_KEY}",
        api: "openai-completions",
        timeoutSeconds: 300, // Optional: extend request timeout for slow local models
        models: [
          {
            id: "your-model-id",
            name: "Local vLLM Model",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 8192,
          },
        ],
      },
    },
  },
}

To keep the provider dynamic without listing every model, add a wildcard to the visible model catalog:

{
  agents: {
    defaults: {
      models: {
        "vllm/*": {},
      },
    },
  },
}

Advanced configuration

Proxy-style behavior

vLLM is treated as a proxy-style OpenAI-compatible /v1 backend, not a native OpenAI endpoint:

Qwen thinking controls

For Qwen models, set compat.thinkingFormat: "qwen-chat-template" on the model row when the server expects Qwen chat-template kwargs. These models expose a binary /think profile (off, on) because Qwen chat-template thinking is an on/off flag, not an OpenAI-style effort ladder.

{
  models: {
    providers: {
      vllm: {
        models: [
          {
            id: "Qwen/Qwen3-8B",
            name: "Qwen3 8B",
            reasoning: true,
            compat: { thinkingFormat: "qwen-chat-template" },
          },
        ],
      },
    },
  },
}

OpenClaw maps /think off to:

{
  "chat_template_kwargs": {
    "enable_thinking": false,
    "preserve_thinking": true
  }
}

Non-off thinking levels send enable_thinking: true. If your endpoint expects DashScope-style top-level flags instead, use compat.thinkingFormat: "qwen" to send enable_thinking at the request root.

Nemotron 3 thinking controls

For vllm/nemotron-3-* models with thinking off, the bundled plugin sends:

{
  "chat_template_kwargs": {
    "enable_thinking": false,
    "force_nonempty_content": true
  }
}

To customize these values, set chat_template_kwargs under the model params. If you also set params.extra_body.chat_template_kwargs, that value wins because extra_body is the last request-body override.

{
  agents: {
    defaults: {
      models: {
        "vllm/nemotron-3-super": {
          params: {
            chat_template_kwargs: {
              enable_thinking: false,
              force_nonempty_content: true,
            },
          },
        },
      },
    },
  },
}

Qwen tool calls appear as text

First confirm vLLM was started with the right tool-call parser and chat template for the model. vLLM documents hermes for Qwen2.5 models and qwen3_xml for Qwen3-Coder models.

Symptoms: skills/tools never run, the assistant prints raw JSON/XML such as {"name":"read","arguments":...}, or vLLM returns an empty tool_calls array when OpenClaw sends tool_choice: "auto".

Some Qwen/vLLM combinations return structured tool calls only when the request uses tool_choice: "required". Force it per model with params.extra_body:

{
  agents: {
    defaults: {
      models: {
        "vllm/Qwen-Qwen2.5-Coder-32B-Instruct": {
          params: {
            extra_body: {
              tool_choice: "required",
            },
          },
        },
      },
    },
  },
}

Replace the model id with the exact id from openclaw models list --provider vllm, or apply the same override from the CLI:

openclaw config set agents.defaults.models '{"vllm/Qwen-Qwen2.5-Coder-32B-Instruct":{"params":{"extra_body":{"tool_choice":"required"}}}}' --strict-json --merge

This is an opt-in workaround: it forces every turn with tools to make a tool call, so use it only for a dedicated model entry where that is acceptable. Do not set it as a global default for all vLLM models, and do not pair it with a proxy that converts arbitrary assistant text into executable tool calls.

Custom base URL

If your vLLM server runs on a non-default host or port, set baseUrl in the explicit provider config:

{
  models: {
    providers: {
      vllm: {
        baseUrl: "http://192.168.1.50:9000/v1",
        apiKey: "${VLLM_API_KEY}",
        api: "openai-completions",
        timeoutSeconds: 300,
        models: [
          {
            id: "my-custom-model",
            name: "Remote vLLM Model",
            reasoning: false,
            input: ["text"],
            contextWindow: 64000,
            maxTokens: 4096,
          },
        ],
      },
    },
  },
}

Troubleshooting

Slow first response or remote server timeout

For large local models, remote LAN hosts, or tailnet links, set a provider-scoped request timeout:

{
  models: {
    providers: {
      vllm: {
        baseUrl: "http://192.168.1.50:8000/v1",
        apiKey: "${VLLM_API_KEY}",
        api: "openai-completions",
        timeoutSeconds: 300,
        models: [{ id: "your-model-id", name: "Local vLLM Model" }],
      },
    },
  },
}

timeoutSeconds applies to vLLM model HTTP requests only: connection setup, response headers, body streaming, and the total guarded-fetch abort. It also raises the LLM idle/stream watchdog ceiling above the implicit ~120s default for this provider. Prefer this over increasing agents.defaults.timeoutSeconds, which controls the whole agent run.

Server not reachable

Check that the vLLM server is running and accessible:

curl http://127.0.0.1:8000/v1/models

If you see a connection error, verify the host, port, and that vLLM started in OpenAI-compatible server mode. OpenClaw trusts the exact configured models.providers.vllm.baseUrl origin for guarded model requests on loopback, LAN, and Tailscale endpoints. Metadata/link-local origins remain blocked without explicit opt-in. Set models.providers.vllm.request.allowPrivateNetwork: true only when vLLM requests must reach another private origin, or false to opt out of exact-origin trust.

Auth errors on requests

If requests fail with auth errors, set a real VLLM_API_KEY that matches your server configuration, or configure the provider explicitly under models.providers.vllm.

Tip

If your vLLM server does not enforce auth, any non-empty value for VLLM_API_KEY works as an opt-in signal for OpenClaw.

No models discovered

Auto-discovery requires VLLM_API_KEY to be set. If you have defined models.providers.vllm, OpenClaw uses only your declared models unless agents.defaults.models includes "vllm/*": {}.

Tools render as raw text

If a Qwen model prints JSON/XML tool syntax instead of executing a skill:

• Start vLLM with the correct parser/template for that model.
• Confirm the exact model id with openclaw models list --provider vllm.
• Add a dedicated per-model params.extra_body.tool_choice: "required" override only if tool_choice: "auto" still returns empty or text-only tool calls.

Warning

More help: Troubleshooting and FAQ.

Related

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

• OpenAI, Native OpenAI provider and OpenAI-compatible route behavior.

• OAuth and auth, Auth details and credential reuse rules.

• Troubleshooting, Common issues and how to resolve them.