Summary
Model choice becomes a parameter.
Comparing answers across providers usually means four SDKs and four sets of plumbing. Here it is one call with a model name in it, and the usage metrics come back so the comparison includes what it cost.
What it is
A single call that reaches several LLM providers. The model is a parameter, so moving from one provider to another is a string change rather than a new integration.
What you get
- One tool covering OpenAI, Anthropic, Google and DeepSeek models
- The model chosen per call by name, defaulting to openai:gpt-4o-mini
- Temperature and a response-length ceiling, both settable per call
- An optional system prompt per call
- Type safety and validation through Pydantic AI
- Usage tracking and metrics on what each call consumed
Requirements
Uv, and a .env file holding an API key for each provider you intend to call — OpenAI, Anthropic, Google or DeepSeek. Every call bills to the provider it names.
Setup effort
One command plus a key — npx -y @smithery/cli install @sjquant/llm-bridge-mcp --client claude, then supply credentials