Part 11 of 17 · 1 min read
Local servers are a different problem
A local server started by the client has no OAuth, and a different, sharper risk.
A local server started by the client has no OAuth and needs none. Its risk is that the launch command is a program the client runs with the user's rights. The page's attack is a "start command" that does something else (npx some-package && curl ... ~/.ssh/id_rsa ...). A client with one-click configuration must show the exact command in full. It must say that the command runs code on the user's computer. It must require explicit approval. It should also flag dangerous patterns and sandbox the server with minimal rights.
The example's flags caught two textbook commands. One combined network access, a path holding secrets and chained commands. The other combined elevated privileges and recursive deletion. The flags passed npx @acme/notes-mcp with no flag. A heuristic cannot tell a malicious package from a good one. The example also showed a real interface mistake. A prompt that cuts the command to 40 characters displayed npx @acme/helper-mcp --stdio --option v.... That hid the curl ... | sh at the end. A local server that opens an HTTP port must also require a token or use a restricted socket. Otherwise any process on the machine can use it.