OpenClaw

• openclaw dashboard 打开Web UI
• openclaw gateway 动gateway
• openclaw gateway start 后台启动gateway、启动任务计划程序（windows）
  • 执行.openclaw/gateway.cmd
• 配置文件 .openclaw/openclaw.json
  • 配置参考
• 环境变量
  • OPENCLAW_HOEM：默认 $HOME，仅限于用在~/.openclaw/, agent dirs, sessions, credentials
    • 基本设置了这个，~/.openclaw/ 仍然会被openclaw的某些功能使用，而且很多外部代码（插件）可能更加不会遵循，比如QQ机器人
  • OPENCLAW_STATE_DIR：默认~/.openclaw
    • 如果设置了OPENCLAW_HOEM，最好设置OPENCLAW_STATE_DIR，因为openclaw有时候不遵循OPENCLAW_HOEM，比如gateway服务的脚本路径 ~/.openclaw/gateway.cmd
  • OPENCLAW_CONFIG_PATH：配置文件~/.openclaw/openclaw.json，遵循OPENCLAW_HOEM
• WebUI：http://127.0.0.1:{gateway.port:18789}/\#token={token}

常见配置需求

Channel

• 模型的处理速度会很大程度上影响回复速度，比如nvidia/deepseek-ai/deepseek-v3.2，几分钟后才回复

Model

• 需要配置models（模型配置）、agents（智能体配置，比如聊天界面、默认智能体）
  • model引用格式：{provider}/{modelId}，如nvidia/deepseek-ai/deepseek-v3.2
  • models配置会被缓存到.openclaw\agents\main\agent\models.json（有时候没有正确同步，比如删除无用的）
• cost：token美元计费，免费推荐都设置为0（仅显示token消耗）
  

  {
    "models": {
      "mode": "merge",
      "providers": {
        "deepseek": {
          "baseUrl": "https://api.deepseek.com",
          "apiKey": "sk-xxxxxx",
          "api": "openai-completions",
          "models": [
            {
              "id": "deepseek-chat",
              "name": "DeepSeek Chat",
              "api": "openai-completions",
              "reasoning": false,
              "input": ["text"],
              "cost": {
                "input": 0.28,
                "output": 0.42,
                "cacheRead": 0.028,
                "cacheWrite": 0
              },
              "contextWindow": 131072,
              "maxTokens": 8192,
              "compat": {
                "supportsUsageInStreaming": true
              }
            },
            {
              "id": "deepseek-reasoner",
              "name": "DeepSeek Reasoner",
              "api": "openai-completions",
              "reasoning": true,
              "input": ["text"],
              "cost": {
                "input": 0.28,
                "output": 0.42,
                "cacheRead": 0.028,
                "cacheWrite": 0
              },
              "contextWindow": 131072,
              "maxTokens": 65536,
              "compat": {
                "supportsUsageInStreaming": true
              }
            }
          ]
        },
        "nvidia": {
          "baseUrl": "https://integrate.api.nvidia.com/v1",
          "apiKey": "nvapi-xxxxx",
          "api": "openai-completions",
          "models": [
            {
              "id": "deepseek-ai/deepseek-v3.2",
              "name": "Nvidia Deepseek-v3.2",
              "api": "openai-completions",
              "reasoning": false,
              "input": ["text"],
              "contextWindow": 131072,
              "maxTokens": 4096,
              "compat": {
                "supportsUsageInStreaming": true
              }
            }
          ]
        }
      }
    },
    "agents": {
      "defaults": {
        "model": {
          "primary": "nvidia/deepseek-ai/deepseek-v3.2"
        },
        "models": {
          "deepseek/deepseek-chat": {
            "alias": "DeepSeek"
          },
          "deepseek/deepseek-reasoner": {
            "alias": "DeepSeek Reasoner"
          },
          "nvidia/deepseek-ai/deepseek-v3.2": {
            "alias": "Nvidia Deepseek-v3.2"
          }
        },
        "workspace": "F:\\data\\.openclaw\\workspace",
        "compaction": {
          "mode": "safeguard"
        }
      }
    }
  }
  

仅本机

{
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback",
    "auth": {
      "mode": "token",
      "token": "xxxxxxxxxxxxx"
    }
  }
}

局域网

{
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "lan",
    "controlUi": {
      "allowedOrigins": ["https://192.168.0.106:18789"],
      // 禁用设备认证（否则需要一个一个openclaw devices approve --latest）
      "dangerouslyDisableDeviceAuth": true
    },
    "auth": {
      "mode": "token",
      "token": "xxxxxxxxxxxxx"
    },
    "tls": {
      "enabled": true,
      "autoGenerate": true, 
      "caPath": "F:\\data\\.openclaw\\openclaw.ca.crt",
      "certPath": "F:\\data\\.openclaw\\openclaw.crt",
      "keyPath": "F:\\data\\.openclaw\\openclaw.key"
    }
  }
}

• autoGenerate：自动生成
  • 或者用https://github.com/FiloSottile/mkcert，生成+自动安装根证书
  • 需要安装openssl
  • 默认生成路径为gateway\tls，注意，不会生成caPath
• 如果想不显示证书不可信或者用TUI：安装caPath
  • TUI的nodejs会进行校验，如果不安装，需要设置环境变量NODE_TLS_REJECT_UNAUTHORIZED=0

问题

“HTTP 401: Authentication Fails, Your api key: *** is invalid”

• 解决方法：删除.openclaw\agents\main\agent\auth-profiles.json
• 从解决401来看，OpenClaw目前在配置修改功能上有很大问题，提供多种配置方式，同时又将配置拆分到不同持久化位置，且同步机制没有做好
• openclaw.json中也包含了auth.profiles，不知道是不是同一个东西，删除后重新运行gateway没有再创建这个auth-profiles.json
• .openclaw\agents\main\agent\models.json 也有
• 好像是初始化程序+配置优先级问题，最好删了，只用 openclaw.json
• 优先级：
  • agents/main/agent/models.json
  • agents/main/agent/auth-profiles.json
  • `openclaw.json``