agentcents MCP
v1.1.3io.github.labham-llc/agentcents
Local LLM cost proxy: forecasts each call and recommends the cheapest model that clears the task.
transport:stdio
runtime:pypi
Target client
run in your project directory
claude mcp add agentcents -- uvx agentcents
Adds it for this project only. Append --scope user to make it available everywhere. Claude Code docs
This listing does not declare its tools. Connect the server and your client will discover them on the handshake.