[Integration Proposal] WhichModel MCP — real-time model pricing plugin for cost-aware AI orchestration

[simonamsharp]

Hi team! 👋

I'd like to propose adding WhichModel as an example MCP plugin for Semantic Kernel. WhichModel is a free MCP server providing real-time pricing and capability data for 100+ LLM models, updated every 4 hours.

Why this matters for Semantic Kernel users

• Agents can check model pricing before making API calls, enabling cost-aware model selection
• Compare models on price, capabilities, and context window in real time
• Track pricing changes across providers (pricing updates every 4 hours)

How it works

WhichModel exposes 4 MCP tools via the Model Context Protocol: recommend_model, compare_models, get_pricing, check_price_changes

Since Semantic Kernel supports MCP, users can add WhichModel as a plugin:

MCP endpoint: https://whichmodel.dev/mcp

What I'm proposing

1. Create a WhichModel plugin example for Semantic Kernel docs
2. Add to the community plugins or MCP examples section

Happy to contribute a PR.

• Website: https://whichmodel.dev
• GitHub: https://github.com/Which-Model/whichmodel-mcp
• MCP endpoint: https://whichmodel.dev/mcp

Would this be valuable for the Semantic Kernel community?

[simonamsharp]

Hi — following up on this integration proposal. WhichModel could serve as a model selection service within Semantic Kernel's connector architecture, enabling cost-aware routing across Azure OpenAI, OpenAI, and other providers.

We've been actively building out integrations since opening this issue — LangChain package is live, Docker deployment ready, and we're listed on major MCP directories.

Happy to contribute a connector implementation if the team is interested.

[simonamsharp]

Hi — second follow-up on this integration proposal. A few updates since last time:

• Our langchain-whichmodel package is live on PyPI
• We have a remote MCP endpoint (npx @anthropic-ai/whichmodel-mcp) ready to use today
• Production Docker images are available

For Semantic Kernel specifically, WhichModel could serve as a model selection service within the connector architecture, enabling cost-aware routing across Azure OpenAI, OpenAI, and other providers. Happy to contribute a PR for a connector or plugin.

Thanks!