What
A can_use_tool request is how Claude Code asks for permission before a tool runs, through the control protocol that SDK clients use. When such a request includes a plugin field naming a plugin, that name is now passed through to the permission handling. Requests that are forwarded on also have their contents adjusted before they are sent.
Why
A permission prompt that comes from a plugin can be traced back to that plugin, so the prompt can say where it came from.
Names in the bundlecan_use_tool
Documented inapi/managed-agents/migration
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed since
Anthropic's documentation has since written up can_use_tool, on Agent SDK reference - Python.
> Commands running with `dangerouslyDisableSandbox: True` have full system access. Ensure your `can_use_tool` handler validates these requests carefully.agent-sdk/python see the edit
One source agreesOne thing we can check says the same as this entry.
Anthropic's documentation agrees
Anthropic's documentation has since written up can_use_tool, on Agent SDK reference - Python.