litellm

npx skills add majiayu000/claude-skill-registry --skill litellm-terminalskills-skills --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

# LiteLLM ## Overview LiteLLM provides a single API to call 100+ LLM providers — OpenAI, Anthropic, Google Gemini, Mistral, Cohere, Azure, Bedrock, Ollama, and more. Write your code once using the OpenAI SDK format, then switch providers by changing a model string. As a proxy server, it adds load balancing, fallbacks, rate limiting, spend tracking, and API key management for teams. ## When to Use - Using multiple LLM providers and want a unified interface - Need automatic fallbacks (if Claude is down, use GPT) - Cost tracking across multiple providers and teams - Load balancing requests across multiple API keys or models - Self-hosted proxy to manage LLM access for a team ## Instructions ### Setup ```bash pip install litellm # Or run as proxy server pip install 'litellm[proxy]' ``` ### SDK Usage (Python) ```python # llm.py — Call any LLM with the same interface from litellm import completion # OpenAI response = completion( model="gpt-4o", messages=[{"role": "user", "content": "Hello!"}], ) # Anthropic — same interface, just change the model string response = completion( model="claude-sonnet-4-20250514", messages=[{"role": "user", "content": "Hello!"}], ) # Google Gemini response =

pip install litellm
Or run as proxy server
pip install 'litellm[proxy]'
Start proxy
litellm --config litellm_config.yaml --port 4000
Call via OpenAI SDK (any language!)
curl http://localhost:4000/v1/chat/completions \