OpenClaw Provider Setup: Anthropic, Ollama, OpenAI, and Gemini

# Option A: Via CLI (recommended)
openclaw config set models.anthropic.api_key sk-ant-your-key-here
openclaw config set models.anthropic.provider anthropic
openclaw config set models.anthropic.model claude-sonnet-4-20250514

# Option B: Via environment variable
export ANTHROPIC_API_KEY=sk-ant-your-key-here
# Add to ~/.zshrc or ~/.bashrc for persistence

# Verify the key works
openclaw models test anthropic

{
  "models": {
    "claude-sonnet": {
      "provider": "anthropic",
      "model": "claude-sonnet-4-20250514",
      "api_key": "sk-ant-...",
      "max_tokens": 4096,
      "temperature": 0.7
    },
    "claude-haiku": {
      "provider": "anthropic",
      "model": "claude-haiku-3-5-20241022",
      "api_key": "sk-ant-...",
      "max_tokens": 2048,
      "temperature": 0.5
    }
  },
  "default_model": "claude-sonnet"
}

---
model: claude-sonnet-4-20250514
provider: anthropic
---

# Orion — Project Manager

## Identity
You are Orion, a senior project manager who keeps teams on track.
You are direct, data-driven, and always focused on shipping.

## Rules
- Open every response with the most important status update
- Flag blockers before they become problems
- Keep responses under 200 words unless asked for detail
- Always suggest a next action at the end of each message

# Set your OpenAI API key
openclaw config set models.openai.api_key sk-your-openai-key
openclaw config set models.openai.provider openai
openclaw config set models.openai.model gpt-4o

# Or via environment variable
export OPENAI_API_KEY=sk-your-openai-key

# Test connectivity
openclaw models test openai

{
  "models": {
    "gpt4o": {
      "provider": "openai",
      "model": "gpt-4o",
      "api_key": "sk-...",
      "max_tokens": 4096,
      "temperature": 0.8
    },
    "gpt4o-mini": {
      "provider": "openai",
      "model": "gpt-4o-mini",
      "api_key": "sk-...",
      "max_tokens": 2048,
      "temperature": 0.5
    }
  }
}

# macOS / Linux: one-line install
curl -fsSL https://ollama.com/install.sh | sh

# Windows: download from ollama.com/download

# Verify Ollama is running
ollama --version

# Pull models for your agents
ollama pull llama3.1        # Best general-purpose (4.7 GB)
ollama pull mistral         # Best for writing agents (4.1 GB)
ollama pull codegemma       # Best for coding agents (4.8 GB)
ollama pull phi3            # Fastest / lightest (2.3 GB)

# Verify the local API is accessible
curl http://localhost:11434/api/tags

{
  "models": {
    "ollama-llama": {
      "provider": "ollama",
      "endpoint": "http://localhost:11434",
      "model": "llama3.1",
      "temperature": 0.7,
      "context_length": 8192,
      "timeout": 120
    },
    "ollama-mistral": {
      "provider": "ollama",
      "endpoint": "http://localhost:11434",
      "model": "mistral",
      "temperature": 0.8
    }
  },
  "default_model": "ollama-llama"
}

# Verify OpenClaw can reach Ollama
openclaw models test ollama-llama

# Assign Ollama to a specific agent
openclaw agents update support-bot --model ollama-llama

# Run your agent
openclaw agent --agent support-bot --message "Hello, are you running locally?"

# Check which model each agent uses
openclaw agents list

# Get your API key from aistudio.google.com/app/apikey
# Then configure OpenClaw
openclaw config set models.gemini.api_key AIza-your-key-here
openclaw config set models.gemini.provider google
openclaw config set models.gemini.model gemini-1.5-pro

# Or via environment variable
export GOOGLE_AI_API_KEY=AIza-your-key-here

# Test the connection
openclaw models test gemini

{
  "models": {
    "gemini-pro": {
      "provider": "google",
      "model": "gemini-1.5-pro",
      "api_key": "AIza-...",
      "max_tokens": 8192,
      "temperature": 0.7
    },
    "gemini-flash": {
      "provider": "google",
      "model": "gemini-1.5-flash",
      "api_key": "AIza-...",
      "max_tokens": 4096,
      "temperature": 0.5
    }
  }
}

# Add to your shell profile for persistence
export ANTHROPIC_API_KEY="sk-ant-your-key"
export OPENAI_API_KEY="sk-your-openai-key"
export GOOGLE_AI_API_KEY="AIza-your-gemini-key"

# Reload your shell
source ~/.zshrc

# OpenClaw reads these automatically if config.json
# references them or if no key is set in config

{
  "models": {
    "claude-sonnet": {
      "provider": "anthropic",
      "model": "claude-sonnet-4-20250514",
      "api_key": "{process.env.ANTHROPIC_API_KEY{'}'}",
      "max_tokens": 4096
    },
    "gpt4o": {
      "provider": "openai",
      "model": "gpt-4o",
      "api_key": "{process.env.OPENAI_API_KEY{'}'}",
      "max_tokens": 4096
    }
  }
}

# See your agent's current model
openclaw agents list

# Switch from Anthropic to Ollama (no SOUL.md changes needed)
openclaw agents update orion --model ollama-llama

# Switch back to Claude Sonnet
openclaw agents update orion --model claude-sonnet

# Switch all agents to a different provider at once
openclaw config set default_model claude-haiku

# Test after switching to confirm behavior
openclaw agent --agent orion --message "What is your current task queue?"