Unclear Only the setting's description is known to have changed; the code that enforces it was not examined.
The disableClaudeAiMcp setting turns off MCP servers that come from claude.ai. MCP (Model Context Protocol) is the standard way outside programs offer Claude extra tools. Its description now says that a claude.ai proxy server you pass in yourself does not connect either, for example one given through --mcp-config or the SDK mcpServers option.
Before, the setting only blocked connectors fetched automatically from claude.ai, and servers you passed in yourself went through the normal MCP trust steps.
If you or your organisation use this setting to keep claude.ai servers out, it now covers servers added by hand as well, not only the automatic ones.
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.
| `false` | The tool stays deferred when their configuration sets `"alwaysLoad": true`. This applies when your server is passed with [`--mcp-config`](/docs/en/cli-reference#cli-flags), supplied by an [Agent SDK application](/docs/en/agent-…mcp see the edit
Only the setting's description is known to have changed; the code that enforces it was not examined.
Anthropic's documentation has since written up --mcp-config, on Connect Claude Code to tools via MCP.