MCP client- isolation of potentially compromised servers?

[danbri]

Does the mcp client allow for some kind of easy (even default) isolation between servers so that malicious content, prompt injecting descriptions etc cannot easily trigger client to perform exfiltration or other regrettable actions?

[danbri]

See also https://x.com/lbeurerkellner/status/1909880854609805545?s=46&t=SzYyMRF8k2fOHaqaxvVzVg