What
An MCP server is an outside tool or service that Claude Code connects to. Some ask you to sign in again partway through, to grant more access than you gave at first. When that happens, a message in the session now says your browser is opening. It shows the sign-in link when that is safe to display. Otherwise it tells you to run /mcp if the browser does not open, because the sign-in screen there shows the link. The message is not saved with the conversation. Before, the link only went to the debug log.
Why
The browser window used to pop up with no explanation in the session.
Names in the bundle/mcp
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.
Added since
A small documentation edit on Configure permissions touched a line naming /mcp after this was published.
| A [`headersHelper`](/docs/en/mcp#trust-a-folder-before-its-headershelper-runs) on a server in `.mcp.json` | Not run until you accept the trust dialog, which appears again naming where the helper is declared. Claude Code connects the serv…permissions see the edit
Confirmed since
Anthropic's documentation has since written up /mcp, on spawn.sh: called once per claimed work item.
The SDK's [Client-side MCP helpers](https://platform.claude.com/docs/en/agents-and-tools/mcp-connector#client-side-mcp-helpers) convert the server's tools into the runnable tools the worker accepts. Install an MCP SDK alongside the Anthrop…managed-agents/self-hosted-sandboxes see the edit
Two sources agreeTwo things we can check say the same as this entry.
Anthropic's documentation agrees
Anthropic's documentation has since written up /mcp, on spawn.sh: called once per claimed work item.
Anthropic's release notes agree
Added /mcp reconnect all in the interactive terminal to retry every MCP server that failed to connect or needs authentication at once