As far as I can tell (and thanks to the user who did some research here) it does not seem possible to dynamically load environment variables for MCP servers. I can appreciate that this may be due to intentional isolation from the workspace, but if so it would be great to have the separate ability to run a bootstrap script in the context of the current workspace (e.g. accessing the current python virtual environment to run a python script that can load environment variables via dotenv) and add an expression(s) for watch files that would re-run the script if eligible files have changed. This is table stakes for being able to work collaboratively with MCP servers. I shouldn’t have to store API keys in my .zshrc, .bashrc, etc… files and I want to be able to commit and share my .cursor/mcp.json files with my team.