"author:AgentMing" matching MCP servers:

Enables MCP clients to send raw text prompts to inexpensive LLM models via the Devin/Windsurf backend, returning answers in result files to avoid context bloat, with support for multi-turn sessions and model discovery.