12+ Providers: Anthropic, OpenAI, Google, Bedrock and Local Models

Chapter 11: Connecting 12+ Providers — Complete Configuration Guide from Anthropic to Local Models

Overview

OpenClaw natively supports 12 official LLM Providers, with an additional 30+ third-party channels available through the plugin system. Whether you're working with top-tier cloud models or locally-deployed open-source models, OpenClaw offers a unified configuration interface. This chapter covers the authentication method, configuration examples, and representative model recommendations for each Provider.

11.1 Model ID Format Specification

OpenClaw uses the unified provider/model-id format to reference all models. This design lets you switch seamlessly between different Providers without modifying business logic code.

Format: <provider>/<model-id>
Examples:
  anthropic/claude-opus-4-6
  openai/gpt-5.5
  google/gemini-3.1-pro
  ollama/llama3.2
  deepseek/deepseek-r1

Reference example in configuration file:

{
  "model": "anthropic/claude-sonnet-4-6",
  "fallback_model": "openai/gpt-5.4-mini"
}

Format Rules

• Provider names are all lowercase, using the officially registered name
• model-id matches the Provider's official API documentation
• Local models use the ollama/ or lmstudio/ prefix

11.2 Anthropic

Authentication

Anthropic uses API Key authentication, passed via the ANTHROPIC_API_KEY environment variable.

# Set environment variable
export ANTHROPIC_API_KEY="sk-ant-api03-xxxxxxxxxxxxxxxxxxxxxxxx"

Configuration Example

{
  "providers": {
    "anthropic": {
      "api_key": "${ANTHROPIC_API_KEY}",
      "base_url": "https://api.anthropic.com",
      "version": "2023-06-01",
      "default_model": "anthropic/claude-sonnet-4-6"
    }
  }
}

Representative Models

Advanced Configuration

{
  "providers": {
    "anthropic": {
      "api_key": "${ANTHROPIC_API_KEY}",
      "max_tokens": 8192,
      "timeout_seconds": 120,
      "extra_headers": {
        "anthropic-beta": "interleaved-thinking-2025-05-14"
      }
    }
  }
}

11.3 OpenAI

Authentication

export OPENAI_API_KEY="sk-proj-xxxxxxxxxxxxxxxxxxxxxxxx"
# Optional: specify organization ID
export OPENAI_ORG_ID="org-xxxxxxxxxxxxxxxx"

Configuration Example

{
  "providers": {
    "openai": {
      "api_key": "${OPENAI_API_KEY}",
      "organization": "${OPENAI_ORG_ID}",
      "base_url": "https://api.openai.com/v1",
      "default_model": "openai/gpt-5.5"
    }
  }
}

Representative Models

11.4 Google Gemini

Authentication

export GEMINI_API_KEY="AIzaSy-xxxxxxxxxxxxxxxxxxxxxxxx"

Configuration Example

{
  "providers": {
    "google": {
      "api_key": "${GEMINI_API_KEY}",
      "base_url": "https://generativelanguage.googleapis.com/v1beta",
      "default_model": "google/gemini-3.1-pro"
    }
  }
}

Representative Models

11.5 Amazon Bedrock

Bedrock is AWS's managed model service. Its authentication method differs from a typical API Key — it uses the AWS standard credential system.

Authentication Methods

Method A: Environment Variables (recommended for development)

export AWS_ACCESS_KEY_ID="AKIAIOSFODNN7EXAMPLE"
export AWS_SECRET_ACCESS_KEY="wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
export AWS_DEFAULT_REGION="us-east-1"

Method B: AWS Profile (recommended for multi-account management)

# ~/.aws/credentials
[openclaw-prod]
aws_access_key_id = AKIAIOSFODNN7EXAMPLE
aws_secret_access_key = wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY

# ~/.aws/config
[profile openclaw-prod]
region = us-east-1

Method C: IAM Role (recommended for production)

# EC2/ECS/Lambda automatically uses instance role — no explicit credential config needed

Configuration Example

{
  "providers": {
    "bedrock": {
      "region": "us-east-1",
      "profile": "openclaw-prod",
      "default_model": "bedrock/anthropic.claude-sonnet-4-6-v1:0"
    }
  }
}

Bedrock vs. Direct Anthropic Comparison

11.6 OpenRouter

OpenRouter acts as a multi-Provider proxy, letting you route to 100+ models with a single API Key.

Use Cases

• Quickly testing model performance across different Providers
• Unified billing management (no need to maintain multiple API Keys)
• Accessing models that cannot be used directly in certain regions

Configuration Example

{
  "providers": {
    "openrouter": {
      "api_key": "${OPENROUTER_API_KEY}",
      "base_url": "https://openrouter.ai/api/v1",
      "default_model": "openrouter/anthropic/claude-opus-4-6",
      "extra_headers": {
        "HTTP-Referer": "https://yourapp.com",
        "X-Title": "OpenClaw Agent"
      }
    }
  }
}

Model ID Format

OpenRouter uses a double-nested format:

openrouter/<original-provider>/<model-id>
Examples:
  openrouter/anthropic/claude-opus-4-6
  openrouter/openai/gpt-5.5
  openrouter/google/gemini-3.1-pro
  openrouter/meta-llama/llama-3.1-70b-instruct

11.7 Ollama (Local)

export OLLAMA_BASE_URL="http://localhost:11434"

{
  "providers": {
    "ollama": {
      "base_url": "${OLLAMA_BASE_URL}",
      "default_model": "ollama/llama3.2"
    }
  }
}

See Chapter 13 for detailed integration instructions.

11.8 DeepSeek

export DEEPSEEK_API_KEY="sk-xxxxxxxxxxxxxxxxxxxxxxxx"

{
  "providers": {
    "deepseek": {
      "api_key": "${DEEPSEEK_API_KEY}",
      "base_url": "https://api.deepseek.com/v1",
      "default_model": "deepseek/deepseek-r1"
    }
  }
}

11.9 MiniMax

export MINIMAX_API_KEY="xxxxxxxxxxxxxxxxxxxxxxxx"
export MINIMAX_GROUP_ID="xxxxxxxxxxxxxxxx"

{
  "providers": {
    "minimax": {
      "api_key": "${MINIMAX_API_KEY}",
      "group_id": "${MINIMAX_GROUP_ID}",
      "base_url": "https://api.minimax.chat/v1",
      "default_model": "minimax/abab6.5s-chat"
    }
  }
}

11.10 xAI (Grok)

export XAI_API_KEY="xai-xxxxxxxxxxxxxxxxxxxxxxxx"

{
  "providers": {
    "xai": {
      "api_key": "${XAI_API_KEY}",
      "base_url": "https://api.x.ai/v1",
      "default_model": "xai/grok-3"
    }
  }
}

11.11 Moonshot AI (Kimi)

export MOONSHOT_API_KEY="sk-xxxxxxxxxxxxxxxxxxxxxxxx"

{
  "providers": {
    "moonshot": {
      "api_key": "${MOONSHOT_API_KEY}",
      "base_url": "https://api.moonshot.cn/v1",
      "default_model": "moonshot/moonshot-v1-128k"
    }
  }
}

11.12 Vercel AI Gateway

Vercel AI Gateway acts as a proxy gateway with unified rate limiting and log tracing.

{
  "providers": {
    "vercel": {
      "api_key": "${VERCEL_AI_GATEWAY_TOKEN}",
      "base_url": "https://ai-gateway.vercel.sh/v1",
      "default_model": "vercel/anthropic/claude-sonnet-4-6"
    }
  }
}

11.13 GLM / Zhipu AI

export ZHIPU_API_KEY="xxxxxxxxxxxxxxxxxxxxxxxx"

{
  "providers": {
    "zhipu": {
      "api_key": "${ZHIPU_API_KEY}",
      "base_url": "https://open.bigmodel.cn/api/paas/v4",
      "default_model": "zhipu/glm-4-plus"
    }
  }
}

11.14 30+ Additional Provider Integrations

Through the plugin system, OpenClaw supports many more Providers. Here are several commonly used ones:

Groq (Ultra-Low Latency Inference)

{
  "providers": {
    "groq": {
      "api_key": "${GROQ_API_KEY}",
      "base_url": "https://api.groq.com/openai/v1",
      "default_model": "groq/llama-3.1-70b-versatile"
    }
  }
}

Mistral AI

{
  "providers": {
    "mistral": {
      "api_key": "${MISTRAL_API_KEY}",
      "base_url": "https://api.mistral.ai/v1",
      "default_model": "mistral/mistral-large-latest"
    }
  }
}

Qwen (Tongyi Qianwen)

{
  "providers": {
    "qwen": {
      "api_key": "${DASHSCOPE_API_KEY}",
      "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
      "default_model": "qwen/qwen-max"
    }
  }
}

Doubao (Volcano Engine)

{
  "providers": {
    "doubao": {
      "api_key": "${ARK_API_KEY}",
      "base_url": "https://ark.cn-beijing.volces.com/api/v3",
      "default_model": "doubao/doubao-pro-128k"
    }
  }
}

Complete Provider Support List

11.15 Multi-Provider Coexistence Configuration

Production environments typically require multiple Providers configured simultaneously to enable failover and cost optimization.

{
  "providers": {
    "anthropic": {
      "api_key": "${ANTHROPIC_API_KEY}",
      "default_model": "anthropic/claude-sonnet-4-6"
    },
    "openai": {
      "api_key": "${OPENAI_API_KEY}",
      "default_model": "openai/gpt-5.4-mini"
    },
    "ollama": {
      "base_url": "http://localhost:11434",
      "default_model": "ollama/llama3.2"
    }
  },
  "default_provider": "anthropic",
  "fallback_chain": [
    "anthropic/claude-sonnet-4-6",
    "openai/gpt-5.4-mini",
    "ollama/llama3.2"
  ]
}

11.16 Provider Capability Comparison Overview

Chapter Summary

• OpenClaw supports 12 official Providers, referenced via the unified provider/model-id format
• Most Providers use API Key authentication; AWS Bedrock uses the IAM credential system
• OpenRouter and Vercel AI Gateway are the best choices for multi-Provider proxying
• Local models (Ollama/LM Studio/vLLM) offer data privacy and zero inference cost advantages
• Multiple Providers can be configured simultaneously, with a fallback_chain for high availability

Next chapter will dive deep into advanced model configuration techniques: Key Rotation, Failover mechanisms, and inference depth control.