Summary
Nine model APIs behind one set of tools.
The value is comparison: put a question to every provider in one call and read the answers next to each other instead of pasting between tabs. Tag preferences and the flagship-versus-cheap switch decide which model each provider actually serves it to.
What it is
An MCP server that puts nine large-language-model APIs behind one set of tools inside a coding client, so a prompt can go to one provider or to every provider at once.
What you get
- One tool per provider: call-chatgpt, call-claude, call-gemini, call-grok, call-deepseek, call-kimi, call-perplexity, call-mistral and call-huggingface
- The same prompt sent to every provider in a single call, so the answers land side by side
- Tag-based model preferences across coding, business, reasoning, math, creative and general work
- A cost preference that decides whether a provider serves its flagship model or a cheaper one
- Hugging Face Inference Router access, which brings OpenAI-compatible Hub models under the same call shape
- Prompt logging with history and statistics
Requirements
Node. Install with npm install -g cross-llm-mcp, from source, or through the one-click install link for Cursor.
Setup effort
One command plus a key — npm install -g cross-llm-mcp, then supply credentials