CConnect

The models, their effort levels and the slash commands are read from Claude Code itself, so a new model or a new command shows up without updating the app Typing / opens the command list, tab fills in the closest match before it cycles, and the list stays out of the way while you write the arguments The context ring knows the real window of every model, and the style Claude answers in is chosen next to the model A subagent now closes with what it did — whether it finished, how long it took and what it spent — both while it runs and when the chat is reopened Thinking blocks can report the tokens they spent, off by default and switchable for a single chat from the visibility menu One of your hooks failing now says so in the chat instead of failing in silence Settings gained a Chats section, holding the trash and the days a chat is kept before Claude Code deletes it, which applies to every account at once Deleted chats are kept outside Claude Code's own folder, where its cleanup can no longer reach them Renaming or recolouring a chat no longer sends it to the top of the list, and the list counts only the chats you can actually open Asking about usage warns when Claude says the limit is close instead of at a fixed number, and shows what extra usage has spent when the account has it turned on MCP servers and plugins added on one account are there on all of them, and you choose which built-in tools Claude gets in every chat Smaller fixes across both apps: selected text and its colour, the progress rings and bars, figures that now read 1M instead of 1000K, and a few blocks that could be selected by accident

Write a review