OpenClaw 🦞

OpenClaw is an open-source local AI assistant that interacts with you through WhatsApp, Telegram, Slack, and other messaging apps. It can perform real actions — managing emails, calendars, flight check-ins, browsing the web, and more. It runs entirely locally, so your data never leaves your device.

Why Pair with HaoAI?

100+ models unified access — One API Key for Claude, GPT, Gemini, Qwen, Doubao, and all major models
Flexible agent strategies — Assign different models to different agents: Sonnet for primary, Opus for reasoning, Haiku for lightweight tasks
99.9% SLA — Multi-node redundancy with automatic failover
Cost control — Unified billing dashboard with real-time token consumption monitoring

As an agentic AI assistant, OpenClaw’s per-task token consumption is relatively high (typically 10K-100K+ tokens). We recommend choosing models based on task complexity to avoid unnecessary costs. This guide provides three configuration options for reference.

Installation

macOS / Linux


curl -sSL https://openclaw.ai/install.sh | bash

Windows (PowerShell)


& ([scriptblock]::Create((iwr -useb https://openclaw.ai/install.ps1)))

Quick Start

1. Get an API Key

Go to the HaoAI Console to create an API Key.

2. Run the Setup Wizard

OpenClaw provides an interactive wizard. Run the following command to get started:


openclaw onboard

In the wizard, select Custom Provider and enter the following:

Setting	Value
Provider Type	`anthropic-messages`
Base URL	`https://api.hao.ai/anthropic`
API Key	Your HaoAI API Key
Model	`anthropic/claude-sonnet-4.6`

3. Start and Verify


openclaw start

Send a message to OpenClaw to test connectivity. If you receive a normal response, the setup is complete.

The wizard automatically generates an openclaw.json configuration file. For more fine-grained control, see the full configuration below.

Full Configuration

OpenClaw manages all configuration through ~/.openclaw/openclaw.json (supports JSON5 format with comments).

Provider Configuration

Providers are defined under models.providers. Each provider needs an API protocol type, URL, and key:

~/.openclaw/openclaw.json


{
  "models": {
    "providers": {
      "haoai-anthropic": {
        "baseUrl": "https://api.hao.ai/anthropic",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "anthropic-messages",
        "models": []
      },
      "haoai-openai": {
        "baseUrl": "https://api.hao.ai/v1",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "openai-responses",
        "models": []
      },
      "haoai-gemini": {
        "baseUrl": "https://api.hao.ai/gemini",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "google-generative-ai",
        "models": []
      }
    }
  }
}

apiKey supports ${ENV_VAR} syntax to reference environment variables, avoiding plaintext keys in config files. Three protocols correspond to different model families: Claude uses haoai-anthropic (anthropic-messages), OpenAI uses haoai-openai (openai-responses), Gemini uses haoai-gemini (google-generative-ai). Other models can connect through haoai-openai.

Models Configuration

Models are defined in the models array of their corresponding provider. We provide three configurations based on your needs:

Premium Setup

All-Claude lineup — Maximum capability for demanding scenarios.

openclaw.json — models.providers


{
  "models": {
    "providers": {
      "haoai-anthropic": {
        "baseUrl": "https://api.hao.ai/anthropic",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "anthropic-messages",
        "models": [
          {
            "id": "anthropic/claude-sonnet-4.6",
            "name": "Claude Sonnet 4.5",
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 64000
          },
          {
            "id": "anthropic/claude-opus-4.6",
            "name": "Claude Opus 4.6",
            "reasoning": true,
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 128000
          },
          {
            "id": "anthropic/claude-haiku-4.5",
            "name": "Claude Haiku 4.5",
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 64000
          }
        ]
      }
    }
  }
}

Model	Role	Use Case
`anthropic/claude-sonnet-4.6`	Primary	Daily chat, task execution, code generation
`anthropic/claude-opus-4.6`	Deep reasoning	Complex analysis, long-chain reasoning, research
`anthropic/claude-haiku-4.5`	Fast response	Simple queries, quick replies, lightweight tasks

Balanced Setup

OpenAI Codex combo — Strong coding capabilities with balanced general performance.

openclaw.json — models.providers


{
  "models": {
    "providers": {
      "haoai-openai": {
        "baseUrl": "https://api.hao.ai/v1",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "openai-responses",
        "models": [
          {
            "id": "openai/gpt-5.3-codex",
            "name": "GPT 5.3 Codex",
            "reasoning": true,
            "input": ["text", "image", "audio", "file"],
            "contextWindow": 512000,
            "maxTokens": 128000
          },
          {
            "id": "openai/gpt-5.1-codex-mini",
            "name": "GPT 5.1 Codex Mini",
            "input": ["text", "image", "audio", "file"],
            "contextWindow": 256000,
            "maxTokens": 65536
          }
        ]
      }
    }
  }
}

Model	Role	Use Case
`openai/gpt-5.3-codex`	Primary + reasoning	Daily chat, complex tasks, code refactoring
`openai/gpt-5.1-codex-mini`	Fast response	Lightweight tasks, code completion, quick Q&A

OpenAI models connect through the haoai-openai provider using the openai-responses protocol.

Cost-Effective Setup

Budget-friendly choice — Low token prices, ideal for high-frequency daily use.

openclaw.json — models.providers


{
  "models": {
    "providers": {
      "haoai-openai": {
        "baseUrl": "https://api.hao.ai/v1",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "openai-responses",
        "models": [
          {
            "id": "bailian/qwen3.5-plus",
            "name": "Qwen 3.5 Plus",
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 64000
          },
          {
            "id": "volcengine/doubao-seed-2.0-code",
            "name": "Doubao Seed 2.0 Code",
            "input": ["text", "image"],
            "contextWindow": 256000,
            "maxTokens": 128000
          }
        ]
      }
    }
  }
}

Model	Role	Use Case
`bailian/qwen3.5-plus`	Primary + reasoning	Daily chat, general tasks, Chinese-optimized
`volcengine/doubao-seed-2.0-code`	Code + fast	Code generation, debugging, quick replies

Agents Configuration

OpenClaw uses agents.defaults for global defaults and agents.list to define different agents, each of which can override default settings:

openclaw.json — agents section


{
  "agents": {
    "defaults": {
      "model": {
        "primary": "haoai-anthropic/anthropic/claude-sonnet-4.6",
        "fallbacks": ["haoai-anthropic/anthropic/claude-haiku-4.5"]
      },
      "models": {
        "haoai-anthropic/anthropic/claude-opus-4.6": { "alias": "opus" },
        "haoai-anthropic/anthropic/claude-sonnet-4.6": { "alias": "sonnet" },
        "haoai-anthropic/anthropic/claude-haiku-4.5": { "alias": "haiku" }
      },
      "thinkingDefault": "low",
      "timeoutSeconds": 600,
      "maxConcurrent": 3
    },
    "list": [
      {
        "id": "main",
        "default": true
      },
      {
        "id": "research",
        "model": {
          "primary": "haoai-anthropic/anthropic/claude-opus-4.6"
        }
      },
      {
        "id": "quick",
        "model": {
          "primary": "haoai-anthropic/anthropic/claude-haiku-4.5"
        }
      }
    ]
  }
}

Agent	Model	Purpose
main	`claude-sonnet-4.6` (inherits defaults)	Default agent for all daily tasks
research	`claude-opus-4.6`	Deep research, complex reasoning, long-form analysis
quick	`claude-haiku-4.5`	Simple Q&A, fast responses, low cost

Model references use the format provider-name/model-id (e.g. haoai-anthropic/anthropic/claude-sonnet-4.6). The models field in defaults defines model aliases for quick switching in conversations with /model opus. Agents in list inherit all defaults configuration and only need to override fields that differ.

Full Configuration Example

Here’s a complete openclaw.json combining Provider, Models, and Agents configuration (premium setup):

~/.openclaw/openclaw.json


{
  "models": {
    "providers": {
      "haoai-anthropic": {
        "baseUrl": "https://api.hao.ai/anthropic",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "anthropic-messages",
        "models": [
          {
            "id": "anthropic/claude-sonnet-4.6",
            "name": "Claude Sonnet 4.5",
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 64000
          },
          {
            "id": "anthropic/claude-opus-4.6",
            "name": "Claude Opus 4.6",
            "reasoning": true,
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 128000
          },
          {
            "id": "anthropic/claude-haiku-4.5",
            "name": "Claude Haiku 4.5",
            "input": ["text", "image", "file"],
            "contextWindow": 200000,
            "maxTokens": 64000
          }
        ]
      },
      "haoai-gemini": {
        "baseUrl": "https://api.hao.ai/gemini",
        "apiKey": "${HAOAI_API_KEY}",
        "api": "google-generative-ai",
        "models": [
          {
            "id": "models/google/gemini-2.5-flash",
            "name": "Gemini 2.5 Flash",
            "reasoning": true,
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 64000
          },
          {
            "id": "models/google/gemini-2.5-pro",
            "name": "Gemini 2.5 Pro",
            "reasoning": true,
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 64000
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "haoai-anthropic/anthropic/claude-sonnet-4.6",
        "fallbacks": ["haoai-anthropic/anthropic/claude-haiku-4.5"]
      },
      "models": {
        "haoai-anthropic/anthropic/claude-opus-4.6": { "alias": "opus" },
        "haoai-anthropic/anthropic/claude-sonnet-4.6": { "alias": "sonnet" },
        "haoai-anthropic/anthropic/claude-haiku-4.5": { "alias": "haiku" }
      },
      "thinkingDefault": "low",
      "timeoutSeconds": 600,
      "maxConcurrent": 3
    },
    "list": [
      {
        "id": "main",
        "default": true
      },
      {
        "id": "research",
        "model": {
          "primary": "haoai-anthropic/anthropic/claude-opus-4.6"
        }
      },
      {
        "id": "quick",
        "model": {
          "primary": "haoai-anthropic/anthropic/claude-haiku-4.5"
        }
      }
    ]
  }
}

Use Cases

Once configured, send commands to OpenClaw through your messaging app:

Email management — “Help me organize today’s unread emails by priority”
Schedule management — “Schedule a team meeting tomorrow at 3 PM”
Information retrieval — “Search for recent articles about AI Agents and summarize the key points”
Code assistance — “Help me review this Python code for security issues”
Workflow automation — “Send a daily team report to Slack #general channel every morning at 9 AM”

Troubleshooting

Cannot connect to HaoAI

Verify the baseUrl configuration is correct:

Anthropic protocol: https://api.hao.ai/anthropic
OpenAI-Response protocol: https://api.hao.ai/v1
Gemini protocol: https://api.hao.ai/gemini

Model not found

Verify the model ID format is correct. When defining models in models.providers, use the full ID returned by HaoAI (e.g. anthropic/claude-sonnet-4.6). When referencing models in agents, add the provider name prefix: haoai-anthropic/anthropic/claude-sonnet-4.6.

High token consumption

OpenClaw’s per-task consumption is relatively high. Recommendations:

Use claude-haiku-4.5 or cost-effective models for daily tasks
Only switch to the research agent (using claude-opus-4.6) for complex tasks
Monitor usage in the HaoAI Console

How to quickly switch models

Use aliases in conversations: /model opus, /model sonnet, /model haiku (requires configuring aliases in agents.defaults.models).