Hi everyone,
I am integrating several Figma projects (Back-office, Front-office, and Design System) into my Cursor workflow using the Model Context Protocol (MCP).
Since these environments are distinct and I want to avoid context “pollution” for the AI, I am looking for the best way to structure my mcp_config.json. My goal is to ensure that each project is isolated with its own File Key and potentially different Access Tokens.
Is the following “multi-instance” approach the recommended way to handle this in Cursor?
JSON
{
"mcpServers": {
"figma-backoffice": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-figma"],
"env": {
"FIGMA_PERSONAL_ACCESS_TOKEN": "TOKEN_1",
"FIGMA_FILE_KEYS": "BACKOFFICE_KEY"
}
},
"figma-frontoffice": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-figma"],
"env": {
"FIGMA_PERSONAL_ACCESS_TOKEN": "TOKEN_2",
"FIGMA_FILE_KEYS": "FRONTOFFICE_KEY"
}
}
}
}
Specifically, I have two questions:
-
Context Switching: Will Cursor effectively treat these as two separate tools? Can I explicitly tell the chat to “Query the figma-backoffice server” to avoid it searching through the Front-office files by mistake?
-
Resource Usage: Does running multiple instances of the same MCP server (
@modelcontextprotocol/server-figma) impact performance or cause conflicts in Cursor?
If there is a better way to manage multiple Figma entry points via a single server instance while keeping the tokens/keys separate, I’d love to hear your thoughts.
Thanks for your help!