The MCP server now exposes six new tools for managing log messages: messages_hide, messages_unhide, messages_fix, messages_open, messages_bot, and messages_unbot. These let an AI client hide noisy messages, mark errors as fixed (or reopen them), and flag messages as bot traffic (or undo that), directly from a conversation with Claude, Codex, or another MCP client.
messages_hide, messages_fix, and messages_open accept an optional flag to apply the change to every other message sharing the same grouping hash, while messages_bot and messages_unbot take a list of up to 500 message IDs per call. All six tools only act when a user has explicitly asked for that specific change, never speculatively while an AI is just investigating errors.
The documentation has been updated to include the new tools.