📋 Claude Guide
🧠 Why Use Claude with ClawdBot?
Claude is the recommended AI model for ClawdBot, especially for coding tasks.
Advantages
Best for Coding
Superior code generation and debugging
Long Context
200K tokens - entire codebases
Prompt Caching
90% cost reduction on repeated context
Instruction Following
Excellent at complex multi-step tasks
Claude Models Comparison
| Model | Input Cost | Output Cost | Best For |
|---|---|---|---|
| Claude 3.5 Sonnet | $3/1M tokens | $15/1M tokens | Coding, analysis (Recommended) |
| Claude 3 Opus | $15/1M tokens | $75/1M tokens | Highest quality (expensive) |
| Claude 3 Haiku | $0.25/1M tokens | $1.25/1M tokens | Fast, cheap responses |
🔑 API Setup
Step 1: Get API Key
- Go to Anthropic Console
- Sign up or log in
- Navigate to "API Keys"
- Click "Create Key"
- Name it "ClawdBot" and copy the key
Step 2: Add Credits
Go to "Billing" and add credits ($5 minimum):
- $5 - Good for testing (~500 messages)
- $20 - Moderate use (~2000 messages)
- $50+ - Heavy use
Step 3: Configure ClawdBot
# Method 1: Environment Variable
export ANTHROPIC_API_KEY="sk-ant-api03-xxx"
# Method 2: Config File
# Edit ~/.clawdbot/config.yaml
providers:
anthropic:
apiKey: "sk-ant-api03-xxx"
model: "claude-3-5-sonnet-20241022"
# Method 3: Interactive Setup
clawdbot config set anthropic.apiKey
Step 4: Test Connection
# Test API connection
clawdbot test anthropic
# Send test message
clawdbot chat "Hello Claude!"
🎯 Model Selection
Claude 3.5 Sonnet (Recommended)
providers:
anthropic:
model: "claude-3-5-sonnet-20241022"
maxTokens: 4096
temperature: 0.7
Best for:
- ✅ Software development
- ✅ Code review and debugging
- ✅ Technical writing
- ✅ Data analysis
- ✅ Complex reasoning tasks
Claude 3 Opus (Premium)
providers:
anthropic:
model: "claude-3-opus-20240229"
Best for:
- ✅ Highest quality responses
- ✅ Creative writing
- ✅ Complex analysis
- ⚠️ 5x more expensive than Sonnet
Claude 3 Haiku (Budget)
providers:
anthropic:
model: "claude-3-haiku-20240307"
Best for:
- ✅ Simple questions
- ✅ Quick responses
- ✅ High-volume use cases
- ✅ 12x cheaper than Sonnet
⚙️ Advanced Configuration
Complete Config Example
# ~/.clawdbot/config.yaml
providers:
default: "anthropic"
anthropic:
apiKey: "${ANTHROPIC_API_KEY}"
model: "claude-3-5-sonnet-20241022"
# Generation parameters
maxTokens: 4096
temperature: 0.7
topP: 1.0
topK: 0
# Prompt caching (recommended!)
enableCaching: true
cacheSystemPrompt: true
# Rate limiting
maxRequestsPerMinute: 50
maxTokensPerMinute: 40000
# Retry logic
maxRetries: 3
retryDelay: 1000 # ms
# Timeout
timeout: 60000 # ms
Temperature Settings
| Temperature | Use Case | Behavior |
|---|---|---|
| 0.0 - 0.3 | Code, factual answers | Deterministic, focused |
| 0.4 - 0.7 | General use (default) | Balanced creativity |
| 0.8 - 1.0 | Creative writing | More varied, creative |
System Prompts
providers:
anthropic:
systemPrompt: |
You are ClawdBot, a helpful AI assistant.
Guidelines:
- Be concise but thorough
- For code, always include comments
- Cite sources when possible
- Ask clarifying questions if needed
Capabilities:
- Browse the web
- Execute code
- Read/write files
- Access user's memory
💾 Prompt Caching (Save 90% on Costs!)
Claude's prompt caching can dramatically reduce costs for repeated context.
How It Works
Cache frequently used context (system prompts, documentation, code) and pay only 10% for cached tokens.
Enable Caching
providers:
anthropic:
enableCaching: true
cacheSystemPrompt: true
# Cache threshold (minimum tokens to cache)
cacheThreshold: 1024
# Cache TTL
cacheTTL: 300 # 5 minutes
Cost Comparison
| Scenario | Without Caching | With Caching | Savings |
|---|---|---|---|
| 100 messages with 10K context | $30 | $6 | 80% |
| 1000 messages with 10K context | $300 | $33 | 89% |
What to Cache
- ✅ System prompts
- ✅ Project documentation
- ✅ Code context
- ✅ User preferences
- ❌ User messages (changes every time)
⚡ Optimization Tips
1. Use Appropriate Models
# Route by task type
routing:
rules:
- pattern: "code|debug|refactor"
model: "claude-3-5-sonnet"
- pattern: "quick|simple|yes|no"
model: "claude-3-haiku"
- pattern: "creative|story|poem"
model: "claude-3-opus"
2. Optimize Context Window
# Limit conversation history
providers:
anthropic:
maxHistoryMessages: 20
# Summarize old messages
summarizeAfter: 50
3. Batch Requests
# Process multiple items in one request
You: Analyze these 5 code files:
1. [file1.py]
2. [file2.py]
...
# Instead of 5 separate requests
4. Set Spending Limits
In Anthropic Console → Billing → Set monthly budget alerts
5. Monitor Usage
# View usage stats
clawdbot stats anthropic
# Check current month spending
clawdbot billing anthropic
❓ Claude FAQ
How much does Claude cost for typical use?
With Claude 3.5 Sonnet and moderate use (50 messages/day), expect
$8-12/month. Enable prompt caching to reduce this by 80-90%.
Is Claude better than GPT-4 for coding?
Generally yes. Claude 3.5 Sonnet excels at code generation, debugging, and
understanding complex codebases.
Can I use Claude without API costs?
No, Claude requires API credits. For free AI, use local Ollama models
instead.
What's the context window size?
All Claude 3 models support 200K tokens (~150K words or ~500 pages).
How do I switch between Claude models?
Change the
model parameter in config.yaml or use runtime
switching: clawdbot model set claude-3-haiku