Plenty of note apps have added a chat sidebar. Most of them are a general-purpose language model that happens to be rendered next to your notes and cannot actually see them.
Notez AI chat is the opposite arrangement: a model with 29 tools pointed directly at your workspace. Press ⌘J.
What it can reach
Finding things
list_notes— keyword search plus filters by tag, board column, property value and date; with no filters, what you've been working onsearch_similar_notes— vector search by meaning, not wordingget_calendar— your reminders and date property values between two datessearch_help— the Notez Help Center, for questions about Notez itself
Understanding structure
get_current_workspace— the workspace with its members, tags, properties and board columnsget_note— read a specific noteget_note_children/get_note_ancestors— walk the note treeget_note_links— the notes a note links to and the notes that link to itget_board— a board with its cards grouped by columnlist_tags— every tag with countslist_properties/get_note_properties— your custom property schema and a note's values
Changing things
create_note— a new note, plain text or Markdownrename_note/edit_note_content/replace_note_content— a new title, a changed passage, or a whole new bodyset_note_property— set a custom property valueset_note_tags/set_note_pinned— replace a note's tags, add it to favorites or take it outmove_note/move_note_to_board_column— nest a note under another, or move a card between board columnsdelete_note/restore_note— move a note to the Bin and bring it backcreate_tags/create_property/update_property/create_board_columns/rename_tag_or_board_column— new tags, properties and columns, and new names for them
The chat is context-aware: with a note open, "this note" resolves to it. No copy-pasting the title.
What it will not do
This matters more than the feature list, so here it is plainly. AI chat cannot:
Delete a tag, a property, an option or a board column
Upload a file from your computer
Reach a workspace other than the one you have open
An assistant that can silently rewrite your archive is an assistant you have to audit constantly, so chat's writes leave a trail. Before chat changes a note's text, Notez saves the version it replaces. Chat names a note and asks before deleting it or rewriting its whole body, and a deleted note waits in the Bin for 30 days.
If chat gets an edit wrong, Version history has the note as it was before.
Search that survives forgetting
Search is hybrid: keyword substring matching on titles and content, unioned with semantic vector search. Both run, and results merge.
This is the difference between search that works on day one and search that still works in year three. Keyword search fails exactly when your archive gets valuable, because you no longer remember the words you used. Semantic search finds the note about "the deployment thing that broke over the holidays" even though those words appear nowhere in it.
Press ⌘K, or on the notes page just start typing.
Similar Notes
Every note has a Similar Notes panel showing up to five related notes with a similarity percentage.
This is the feature that surfaces things you were not looking for. You write a note about API rate limiting and Notez quietly points at something you wrote seven months ago about the same problem. No tagging discipline required, no manual linking — the connection is computed from the writing itself.
Under the hood: text-embedding-3-small at 1536 dimensions, chunked, indexed with pgvector HNSW.
Linked notes and the graph
Separately from semantic similarity, you can link notes explicitly. Type [[ in the editor to link an existing note.
The Linked notes panel shows those connections as a list or a graph, each note marked by direction — incoming, outgoing, or both. Backlinks come free: link A to B and B knows about A.
Two systems, doing different jobs. Explicit links capture connections you know about. Similar Notes surfaces the ones you do not.
The limits, stated
Every AI feature has a budget, and vagueness helps no one:
Model:
gpt-6-sol, from OpenAITool rounds: up to 5 per message
Context: the last 10 messages
Rate limit: 10 chat requests per minute
Conversation length: 100 user messages
Retention: conversations are cleaned up after 90 days
Monthly token allowances by plan:
Free | Pro | Ultra | |
|---|---|---|---|
AI chat tokens | 200,000 | 2,000,000 | 20,000,000 |
AI embedding tokens | 500,000 | 5,000,000 | 50,000,000 |
Usage is visible in Settings → Upgrade and billing with progress bars and a reset date, and you get a notification as you approach a limit. Nothing fails silently.
Things worth asking it
"What have I written about focus and deep work?"
"Find all my notes tagged research and summarize the top three."
"Tag everything in my inbox that mentions pricing."
"Move the notes in Doing that I have not touched this month back to Backlog."
"What did I decide about the database migration, and why?"
The last one is the reason any of this exists. Notes are worth keeping only if the reasoning inside them can be recovered later. Storage was never the hard part.
Continue reading
Semantic Search for Notes, Explained — how embeddings find the note that says "memoize" when you search for "caching", and where keyword search still wins.
"What are my tasks today?" — why the chat answered with the wrong list until it could see the board on your screen.
Custom instructions for AI chat: one set for the team, one for you — standing instructions for a whole workspace and for yourself, and why the owner's go last in the prompt.