FAQ: Frontend Review MCP Server

Q: What is an MCP Server? A: MCP (Model Context Protocol) is an open standard that standardizes how applications provide context to LLMs. An MCP server acts as a bridge, allowing AI models to access external data sources and tools.

Q: What does the Frontend Review MCP Server do? A: It performs a visual review of UI edit requests by comparing screenshots of the page before and after the edit, ensuring the changes align with the user’s request.

Q: How do I install the Frontend Review MCP Server? A: You can install it in a Cursor project by adding it to the .cursor/mcp.json file or globally in Cursor settings. For Windsurf, add it to your ~/.codeium/windsurf/mcp_config.json file.

Q: What arguments does the reviewEdit tool require? A: It requires the absolute paths to the “before” and “after” screenshots, along with a detailed description of the UI edit request.

Q: What review models are used by the server? A: The server defaults to Qwen/Qwen2-VL-72B-Instruct and falls back to Qwen/Qwen2-VL-7B-Instruct, meta-llama/Llama-3.2-90B-Vision-Instruct, and mistralai/Pixtral-12B-2409 if the primary model fails.

Q: Can I use a different review model? A: Yes, you can specify a different model using the MODEL argument in the command line.

Q: How do I take screenshots for the review? A: You can use any MCP server to take screenshots. Make sure your AI agent saves the screenshots in a known location (e.g., /screenshots).

Q: What AI instructions should I include in my agent’s prompt? A: Include instructions to take screenshots before and after edits, save them to a specific folder, and use the find command to get the absolute paths for the reviewEdit tool.

Q: What is UBOS? A: UBOS is a full-stack AI Agent Development Platform focused on bringing AI Agents to every business department. Our platform helps you orchestrate AI Agents, connect them with your enterprise data, build custom AI Agents with your LLM model and Multi-Agent Systems.