Coming soon
Nothing on this page exists yet
The items below are planned or under study. They may change shape, name, or order before they arrive. When they arrive, they go into the release notes and the reference.
Under study
These experience changes have already been evaluated, but need a decision before they go in:
- Adopting an Obsidian vault where it is, without cloning it into a new folder, including a folder that is not a git repository yet. Today
initclones the vault into a new folder, and you open that folder in Obsidian (see Adopt an existing vault). - The plugin as the default path in Claude Code, installed by
inititself when Claude Code is on the machine. Todayinitregisters the MCP server directly, and the plugin is installed separately. - Automatic sync turned on by
init. Today it only shows up among the optional steps at the end ofinit,statusanddoctor(cabecao schedule on). - A
cabecao setup: a list of what is on and what is missing (skills, automatic sync, semantic search, Obsidian), with a single question to apply the recommended setup. - CLI messages in English, chosen by the system language. Today the CLI speaks Portuguese; the site and the READMEs are in both languages.
Planned
- Shell autocompletion (
cabecao completion bash|zsh|fish|powershell). - Consolidation scheduled by Cabeção itself:
cabecao consolidate --runalready calls an agent on its own, butcabecao scheduleonly schedules the sync; today, scheduling the consolidation is done with the system scheduler (see Consolidate with an agent). - More agents in
consolidate --run: GitHub Copilot CLI and OpenCode have options to restrict tools, but have not been tested yet. - Delete for good (
cabecao purge): remove a note from the git history, for the rare case of something that should never have been there. Todayforgetarchives, and the archived note stays in git. - Graph explorer inside Claude, to browse notes and relations in a window of the conversation.
- More from the answer provider (
askitself already exists): topic summaries and consolidation help written by the provider, always under the same restrictions as an agent; a model that runs embedded, with no server; and semantic search with Ollama.
Want to follow along or share your opinion? Open an issue.