Skip to main content

API Keys Guide

OpenClaw requires at least one LLM provider's API Key to function. This guide walks you through obtaining API Keys from major LLM providers and configuring them in OpenClaw.

Cost Reminder

Using LLM APIs incurs costs. Each provider has different billing structures — check pricing before signing up. Most providers offer free credits for initial testing.

Quick Comparison​

ProviderRecommended ModelFree TierCost per Million Tokens (approx.)Best For
OpenAIGPT-5.2 CodexLimited$2-15General conversation, code generation
AnthropicClaude Opus 4.6Limited$3-15Long-form analysis, complex reasoning
GoogleGemini 2.5 ProGenerous free tier$1-7Multimodal, cost-effective
DeepSeekDeepSeek-V3Very low cost$0.1-0.5Budget-conscious, Chinese-optimized
Ollama (Local)Various open-source modelsCompletely free$0 (hardware cost only)Offline use, privacy-first

OpenAI API Key​

Steps​

  1. Go to platform.openai.com
  2. Sign up or log in to your existing account
  3. Click API Keys in the left sidebar
  4. Click Create new secret key
  5. Name the key (e.g., openclaw-production)
  6. Copy and securely store the key (shown only once)

OpenClaw Configuration​

llm:
  provider: openai
  api_key: sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  model: gpt-5.2-codex

Or use environment variables:

export OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Cost Management Tips​

  • Set a monthly usage cap (Monthly Budget)
  • Monitor consumption via the Usage Dashboard
  • Use cheaper models (e.g., GPT-4o-mini) during development and testing
API Key Security

Never commit API Keys to a Git repository or share them with others. Use environment variables or a dedicated secrets manager.

Anthropic API Key​

Steps​

  1. Go to console.anthropic.com
  2. Sign up (phone number verification required)
  3. Navigate to Settings > API Keys
  4. Click Create Key
  5. Copy and securely store

OpenClaw Configuration​

llm:
  provider: anthropic
  api_key: sk-ant-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  model: claude-opus-4-6

Or environment variable:

export ANTHROPIC_API_KEY=sk-ant-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Claude models excel at:

  • Long-context understanding: Up to 1M token context window
  • Complex reasoning tasks: Multi-step logical analysis
  • Code review: Precise code analysis and suggestions

Google Gemini API Key​

Steps​

  1. Go to aistudio.google.com
  2. Sign in with your Google account
  3. Click Get API Key
  4. Select or create a Google Cloud Project
  5. Copy the generated API Key

OpenClaw Configuration​

llm:
  provider: google
  api_key: AIzaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  model: gemini-2.5-pro

Key Features​

  • Multimodal support (images, video, audio)
  • Relatively generous free tier
  • Good integration with Google Workspace

DeepSeek API Key​

Steps​

  1. Go to platform.deepseek.com
  2. Register an account (supports mainland China phone numbers)
  3. Enter the console and find the API Keys section
  4. Create a new Key and copy it

OpenClaw Configuration​

llm:
  provider: deepseek
  api_key: sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  model: deepseek-v3

Best For​

  • Budget-limited: Pricing is 1/10 to 1/30 of OpenAI
  • Chinese-language scenarios: Specifically optimized for Chinese
  • Mainland China users: Accessible without a VPN
Mainland China Note

DeepSeek is the preferred LLM provider for mainland China users due to direct access without a VPN and extremely low pricing. Note that state-owned enterprises may have additional compliance requirements for AI services. See Chinese Ecosystem for details.

Ollama (Local Models)​

Installation​

Ollama does not require an API Key — it runs open-source models on your local machine.

  1. Go to ollama.com to download and install
  2. After installation, pull models:
ollama pull llama3.3
ollama pull deepseek-r1:32b
  1. Configure in OpenClaw:
llm:
  provider: ollama
  base_url: http://localhost:11434
  model: llama3.3

Pros & Cons​

ProsCons
Completely freeRequires capable hardware (GPU)
Data never leaves your machineModel capabilities usually lag behind cloud
No internet connection neededLarge models are slow to start
No API rate limitingYou manage model updates

Multi-Model Routing Strategy​

OpenClaw supports configuring multiple LLM providers simultaneously, automatically selecting the best model based on task type:

llm:
  default_provider: anthropic
  default_model: claude-opus-4-6

  routing:
    - task: code_generation
      provider: openai
      model: gpt-5.2-codex
    - task: casual_chat
      provider: deepseek
      model: deepseek-v3
    - task: image_analysis
      provider: google
      model: gemini-2.5-pro
    - task: offline_fallback
      provider: ollama
      model: llama3.3
Cost-Saving Strategy

Route casual conversations to DeepSeek or Ollama, and only use Claude or GPT when high-quality reasoning is needed. This can significantly reduce your monthly API costs.

API Key Security Best Practices​

  1. Use environment variables: Do not write Keys directly in config files
  2. Rotate regularly: Replace API Keys every 90 days
  3. Set usage caps: Configure monthly budgets in each provider's console
  4. Least privilege: If the provider supports it, set minimum necessary permissions for the Key
  5. Monitor abnormal usage: Regularly check API usage dashboards
Leak Response

If your API Key is accidentally leaked (e.g., committed to GitHub):

  1. Immediately revoke the Key in the provider's console
  2. Create a new Key
  3. Check your billing for abnormal charges
  4. Review Git history and clean up sensitive data