Using LM Studio with OpenClaw

Prerequisites

• ✓LM Studio desktop app or `llmster` headless installed on the same machine where you expose the LM Studio HTTP server.
• ✓OpenClaw CLI installed and available as the `openclaw` command.
• ✓Network access from OpenClaw to the LM Studio server URL `http://localhost:1234` (or your chosen host/port).
• ✓An LM Studio API token if authentication is enabled in your LM Studio server configuration.

Steps

1. 1

   Install LM Studio and start the local server

   Install LM Studio (desktop) or the `llmster` headless daemon so you can expose a local HTTP API that OpenClaw talks to. This script comes from LM Studio and sets up the tooling on your machine.

   bashCopy

   curl -fsSL https://lmstudio.ai/install.sh | bash

2. 2

   Run the LM Studio daemon and HTTP server

   Start the LM Studio background daemon and HTTP server so OpenClaw can reach it on a stable port. If you use the desktop app you can start it there; for headless setups, run these commands and keep them running.

   bashCopy

   lms daemon up
   
   lms server start --port 1234

3. 3

   Export your LM Studio API token for OpenClaw

   OpenClaw reads the LM Studio token from the `LM_API_TOKEN` environment variable when you onboard the provider. If your LM Studio server has authentication enabled, set the real token; if auth is disabled, you still need to export any non-empty value.

   bashCopy

   export LM_API_TOKEN="your-lm-studio-api-token"

4. 4

   Use a placeholder token for unauthenticated LM Studio servers

   When LM Studio authentication is disabled, OpenClaw still expects `LM_API_TOKEN` to be set. Export a placeholder value so onboarding and later requests succeed without 401 errors.

   bashCopy

   export LM_API_TOKEN="placeholder-key"

5. 5

   Run OpenClaw onboarding and select LM Studio

   Run the OpenClaw onboarding flow to register LM Studio as a model provider. During the prompts, choose `LM Studio` so OpenClaw writes the correct provider config and auth profile.

   bashCopy

   openclaw onboard

6. 6

   Set your default LM Studio model in OpenClaw

   After onboarding, point OpenClaw at the specific LM Studio model you want as the default. Use the `openclaw models set` command with the `lmstudio/` prefix plus the LM Studio model key.

   bashCopy

   openclaw models set lmstudio/qwen/qwen3.5-9b

7. 7

   Script non-interactive onboarding for CI or provisioning

   For automated setups, use non-interactive onboarding so you can configure LM Studio without prompts. This variant assumes `LM_API_TOKEN` is already exported and writes the LM Studio provider and auth profile into your config.

   bashCopy

   openclaw onboard \
     --non-interactive \
     --accept-risk \
     --auth-choice lmstudio

8. 8

   Pin base URL, API key, and model in non-interactive onboarding

   If you need to override the default LM Studio URL or specify the model and API key explicitly, pass them as flags. This is useful when LM Studio runs on a non-default host/port or when you want a specific model key baked into your config.

   bashCopy

   openclaw onboard \
     --non-interactive \
     --accept-risk \
     --auth-choice lmstudio \
     --custom-base-url http://localhost:1234/v1 \
     --lmstudio-api-key "$LM_API_TOKEN" \
     --custom-model-id qwen/qwen3.5-9b

9. 9

   Configure LM Studio explicitly in your OpenClaw models config

   If you manage OpenClaw config by hand, add an explicit `lmstudio` provider block. This pins the base URL, API key, and one or more LM Studio models with their capabilities and token limits.

   jsonCopy

   {
     models: {
       providers: {
         lmstudio: {
           baseUrl: "http://localhost:1234/v1",
           apiKey: "${LM_API_TOKEN}",
           api: "openai-completions",
           models: [
             {
               id: "qwen/qwen3-coder-next",
               name: "Qwen 3 Coder Next",
               reasoning: false,
               input: ["text"],
               cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
               contextWindow: 128000,
               maxTokens: 8192,
             },
           ],
         },
       },
     },
   }

Configuration

Troubleshooting

# Start via desktop app, or headless:
lms server start --port 1234

curl http://localhost:1234/api/v1/models

Frequently asked questions