"author:sarveshkochhar" matching MCP servers:

Enables Claude to delegate bounded, isolated units of work to external LLMs such as Z.AI GLM or any OpenAI-compatible endpoint, running single or up to 16 concurrent workers in separate tool-less workspaces. Results come back through normalized MCP tools that list models, launch, poll, cancel, and clean up agents without exposing credentials or leaking worker output as instructions.