I built an MCP gateway that lets models use Microsoft Copilot for vision and documents

I built a small MCP project and would love feedback from people using OpenCode, DeepSeek, GLM/Z.ai models, or other coding agents. github.com/yurilopes/Copilot-Tools-Gateway The basic idea is: keep your main coding model as the main agent, but let it call Microsoft Copilot as an auxiliary tool when it needs capabilities the model/tooling may not have, like vision, screenshot understanding, image generation, or document/file-assisted questions. This is especially useful with models like GLM-5.2 or De