LiteLLM Gateway Settings

Provider Configuration

Copy
# Point Claude Code at your LiteLLM gateway
export ANTHROPIC_BASE_URL=http://localhost:4000
export ANTHROPIC_API_KEY=sk-litellm-master-key

Settings File

Copy
{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:4000",
    "ANTHROPIC_API_KEY": "sk-litellm-master-key"
  }
}

LiteLLM Proxy Setup

1. Install LiteLLM

Copy
pip install litellm[proxy]

2. Create Configuration (litellm_config.yaml)

Copy
model_list:
  - model_name: claude-sonnet-4-20250514
    litellm_params:
      model: anthropic/claude-sonnet-4-20250514
      api_key: sk-ant-your-key

  - model_name: claude-haiku-4-20250514
    litellm_params:
      model: anthropic/claude-haiku-4-20250514
      api_key: sk-ant-your-key

general_settings:
  master_key: sk-litellm-master-key
  database_url: sqlite:///litellm.db

litellm_settings:
  max_budget: 100  # Monthly budget cap in USD
  budget_duration: monthly

3. Start the Proxy

Copy
litellm --config litellm_config.yaml --port 4000

Team Features

• Centralized API keys: Team members don't need individual provider keys
• Budget controls: Set per-user and per-team spending limits
• Usage logging: Track token usage and costs per user
• Model routing: Route different models to different providers
• Fallback chains: Auto-failover between providers

Troubleshooting

Issue	Solution
Connection refused	Ensure LiteLLM proxy is running on port 4000
Auth failed	Check master_key matches ANTHROPIC_API_KEY
Model not found	Verify model_name in litellm_config.yaml
Slow responses	Check proxy logs with litellm --debug

Security Notes

• Run the proxy on localhost only, or behind a reverse proxy with TLS
• Use unique API keys per team member for usage tracking
• The proxy stores all API keys - secure the config file appropriately