Collections & documents

Collections

A collection is a named bucket for related work — like a folder. meeting-notes, roadmap, recipes. You and your AI both create them; an AI will make one on the fly when you ask it to "save this to a new collection".

Documents

A document is a single piece of markdown living inside a collection, addressed by a key. Keys are short, URL-friendly handles — standup, 2026-q3-plan — unique within their collection. Writing to an existing key updates that document and bumps its version; writing to a new one creates it.

The body is plain GitHub-flavored markdown: headings, lists, links, code blocks, and task checkboxes. It renders cleanly in the web editor and stays plain text you can export at any time.

Task checkboxes

Any - [ ] line becomes a clickable checkbox in the UI. Your AI can tick or untick it with the toggle_task tool, so a shared checklist stays in sync whether you or Claude touches it.

## Launch
- [x] Write the docs
- [ ] Record the demo
- [ ] Ship it

The round-trip

This is what makes a document a living document, not throwaway output:

1. AI writes a document with put_doc.
2. You edit it in the web editor — fix, extend, reorganize.
3. AI reads it back next session with get_doc and continues.

Finding things again

Ask your AI to list_collections or list_docs to browse, or search_docs to full-text search across everything. In the web UI, the same search lives in the top bar.

Yours to take

Because every document is plain markdown, export is a first-class feature: download a single document, a whole collection as markdown, or your entire workspace as a ZIP. Your work leaves whenever you want it to.

Next