Skip to main content
Weaver provides native SDK support for Anthropic’s Claude models with both API key and OAuth authentication.

​
Authentication Methods

​
1. API Key Authentication

Standard Anthropic API key authentication.

​
Get API Key

  1. Visit Anthropic Console
  2. Navigate to API Keys
  3. Create a new API key
  4. Copy the key (starts with sk-ant-)

​
Configure Weaver

Option A: Configuration File Add to ~/.weaver/config.json: 
{
  "providers": {
    "anthropic": {
      "api_key": "sk-ant-...",
      "api_base": "https://api.anthropic.com/v1"
    }
  },
  "agents": {
    "defaults": {
      "provider": "anthropic",
      "model": "claude-sonnet-4-5-20250929"
    }
  }
}
Option B: Environment Variable Add to .env: 
ANTHROPIC_API_KEY=sk-ant-...

​
2. OAuth Authentication

Use OAuth for token-based authentication: 
weaver auth login --provider anthropic
Configure in ~/.weaver/config.json: 
{
  "providers": {
    "anthropic": {
      "auth_method": "oauth"
    }
  }
}
OAuth authentication automatically refreshes tokens and is recommended for long-running deployments.

​
Available Models

Weaver supports all Claude model variants:
ModelDescriptionContext Window
claude-sonnet-4-5-20250929Latest Sonnet (default)200K tokens
claude-3.5-sonnet-20241022Previous Sonnet version200K tokens
claude-3-opus-20240229Most capable Claude 3200K tokens
claude-3-haiku-20240307Fast and efficient200K tokens

​
Configuration Options

​
api_key
string
Anthropic API key (required for API key auth)
​
api_base
string
default:"https://api.anthropic.com/v1"
Anthropic API endpoint URL
​
auth_method
string
Authentication method: api_key (default), oauth, or token
​
proxy
string
HTTP/HTTPS proxy URL for API requests (optional)

​
Model Parameters

Configure model behavior: 
{
  "agents": {
    "defaults": {
      "model": "claude-sonnet-4-5-20250929",
      "max_tokens": 8192,
      "temperature": 0.7
    }
  }
}
​
max_tokens
integer
default:"4096"
Maximum tokens in response (Claude enforces max_tokens)
​
temperature
float
default:"0.7"
Controls randomness in responses (0.0 = deterministic, 1.0 = very random)

​
Message Format

Weaver automatically converts between standard and Claude message formats:

​
System Messages

{
  "role": "system",
  "content": "You are a helpful assistant"
}
Converted to Claude’s system parameter.

​
User Messages

{
  "role": "user",
  "content": "Hello!"
}

​
Tool Results

{
  "role": "tool",
  "tool_call_id": "call_123",
  "content": "{\"result\": \"success\"}"
}
Converted to Claude’s tool_result blocks. Source: pkg/providers/claude_provider.go:59-97

​
Tool Calling

Claude natively supports tool calling. Weaver automatically converts tool definitions: 
// Weaver format
type ToolDefinition struct {
  Type     string
  Function ToolFunctionDefinition
}

// Converted to Claude ToolParam
{
  "name": "function_name",
  "description": "Function description",
  "input_schema": {
    "properties": {...},
    "required": [...]
  }
}
Source: pkg/providers/claude_provider.go:125-149

​
Tool Response Parsing

Claude returns tool calls in content blocks: 
{
  "type": "tool_use",
  "id": "toolu_123",
  "name": "function_name",
  "input": {...}
}
Weaver parses these into standard ToolCall format. Source: pkg/providers/claude_provider.go:151-194

​
Implementation Details

Weaver’s Anthropic integration uses:
  • ClaudeProvider with native Anthropic SDK (github.com/anthropics/anthropic-sdk-go)
  • Automatic token refresh via token source function
  • Native tool calling support with input schema validation
  • Proper message format conversion for system/user/assistant/tool roles
Source: pkg/providers/claude_provider.go

​
Usage Examples

​
Basic Usage

# Use latest Sonnet
weaver chat --model claude-sonnet-4-5-20250929

# Use Opus for complex tasks
weaver chat --model claude-3-opus-20240229

# Use Haiku for speed
weaver chat --model claude-3-haiku-20240307

​
With Tool Calling

{
  "agents": {
    "defaults": {
      "model": "claude-sonnet-4-5-20250929",
      "max_tokens": 8192,
      "max_tool_iterations": 20
    }
  }
}

​
With Proxy

Configure proxy for restricted networks: 
{
  "providers": {
    "anthropic": {
      "api_key": "sk-ant-...",
      "proxy": "http://proxy.company.com:8080"
    }
  }
}

​
Response Handling

Weaver parses Claude responses with proper finish reason mapping:
Claude ReasonWeaver ReasonDescription
end_turnstopNatural completion
tool_usetool_callsTool calling triggered
max_tokenslengthHit token limit
Source: pkg/providers/claude_provider.go:174-183

​
Troubleshooting

​
API Key Issues

# Verify API key
echo $ANTHROPIC_API_KEY

# Test with curl
curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-3-haiku-20240307","max_tokens":100,"messages":[{"role":"user","content":"Hi"}]}'

​
OAuth Issues

# Re-authenticate
weaver auth logout --provider anthropic
weaver auth login --provider anthropic

# Check credentials
weaver auth list

​
Common Errors

  • Verify API key is correct and starts with sk-ant-
  • For OAuth: Run weaver auth login --provider anthropic
  • Check that the API key hasn’t been revoked
  • Claude requires max_tokens parameter
  • Set in config: "max_tokens": 8192
  • Weaver defaults to 4096 if not specified
  • Check your API usage in Anthropic Console
  • Implement exponential backoff
  • Consider upgrading to a higher tier
  • Ensure tool definitions include proper schema
  • Check that required fields are specified
  • Verify JSON in tool responses is valid

​
Next Steps

Provider Overview

Back to all providers

Tool Calling

Use tools with Claude

Build docs developers (and LLMs) love

Get started for freeTalk to us