Blog

The following is a brief outline about how to install OpenClaw with macOS client device + local LLM instance.

Environment

• LLM server : Nvidia DGX Spark
• LLM Model : GPT OSS 120b
• OpenClaw Client : macMini

Installing GPT OSS 120b on DGX Spark

llama.cpp is significantly faster than ollama for one client system. So, we use llama.cpp as local LLM server.

1. download llama.cpp

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp

2. compile llama.cpp
compilation needs CUDA option for DGX Spark

cmake -B build -DGGML_CUDA=ON
cmake --build build --config Release -j

3. download GPT OSS 120b

build/bin/llama-cli --hf unsloth/gpt-oss-120b-GGUF

Optionally, you can install Qwen3-Coder model also.

build/bin/llama-cli --hf unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF

4. enable llama.cpp as Ubuntu service.

create the following file with the below contents. : /etc/systemd/system/llama-server.service

[Unit]
Description=Llama.cpp API Server
After=network.target

[Service]
Type=simple
User=pinebud
WorkingDirectory=/your/home/directory/llama.cpp
ExecStart=/your/home/directory/llama.cpp/build/bin/llama-server --host your_local_ip_address --port your_desired_local_port --api-key your_desired_key
Restart=always

[Install]
WantedBy=multi-user.target

enable the service

sudo systemctl daemon-reload
sudo systemctl enable llama-server

You can connect to the service through the IP you setup and port number. for ex) http://192.168.1.1:8081

Installing OpenClaw on macMini

1. install homebrew

reference : https://brew.sh/

2. install npm

brew install npm

3. install openclaw

npm i -g openclaw
openclaw onboard

4. configure LLM settings

add the following settings after onboarding setup (~/.openclaw/openclaw.json)

  "models": {
    "providers": {
      "llama.cpp": {
        "baseUrl": "http://your_llm_ip:your_llm_port/v1",
        "apiKey": "your_desired_api_key",
        "api": "openai-completions",
        "models": [
          {
            "id": "unsloth/gpt-oss-120b-GGUF",
            "name": "GPT-OSS",
            "reasoning": true,
            "input": [
              "text"
            ]
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "llama.cpp/unsloth/gpt-oss-120b-GGUF"
      },
      "models": {
        "llama.cpp/unsloth/gpt-oss-120b-GGUF": {
          "alias": "GPT-OSS"
        }
      }
    }
  },

5. test run gateway

by running gateway with –verbose option, you can debug the issue on connection more easily.

openclaw gateway stop
openclaw gateway --verbose

Setting up Slack Bot with OpenClaw

References

https://github.com/ggml-org/llama.cpp

https://huggingface.co/unsloth/gpt-oss-120b-GGUF

https://huggingface.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF

https://openclaw.ai