Tools an AI app can call in MindMax
A connected app sees a set of tools built from the same functions MindMax's own assistant uses. The names below are what appear in its tool list; the wording it reads is what you would read here.
Three access levels run through the list. A read-only connection sees only the first group.
- Read — opens and reads. Changes nothing.
- Propose — produces a proposal or a comment. Your document is untouched until you accept.
- Write — creates something new, or appends. Never overwrites, never deletes.
Read
| Tool | What it returns |
|---|---|
get_active_context | What you are looking at: open workspaces, the current document with its section path, the selection and the passage around it, reading position, open tabs, the canvas in front |
find_documents | Which documents are about something — path, format, size, one-sentence gist — filterable by folder, tag, property or modification date |
search_knowledge | Passages across every file in the workspace, each with its section path, its anchor and a quote handle; exact-wording mode, or restricted to one document or section |
read_document | A short document in full; a long one as a map with section ids and notes; one section read whole; page images for figures; a .pptx by slide |
get_annotations | Comment threads with every message, each marked with its intent and whose turn it is — by default only the ones waiting on the AI — plus highlights with the meaning you gave each colour, bookmarks and reading-lens notes |
get_links | Outgoing links, backlinks and unlinked mentions for a note |
get_document_history | Saved versions, including which were written by you, by MindMax or by a connected app, and what changed between two of them |
get_review_status | What you did with a proposal: pending, accepted, partially accepted (which parts), rejected, withdrawn |
open_in_mindmax | Opens a document and scrolls to a section, a quoted passage or a page, so the app can point at what it means |
read_canvas | A canvas as an outline: cards, who wrote each one, the points under it, the relations drawn between cards |
list_thinking_methods | MindMax's named thinking methods by group, with the card shape each produces; one method in full when named |
list_document_methods | The document jobs MindMax knows how to do (proofreading, review, verification…), so the app can follow one instead of improvising |
check_diagram | Draws a Mermaid or infographic block with the editor's own renderer and returns the syntax error, if any, plus what does not match the cited source |
get_step | One self-contained work package for a running method: the instructions, the input, the expected output shape |
Propose
| Tool | What it does |
|---|---|
suggest_edits | Proposes changes to an existing document: a local replacement, a section rewrite, a whole-document rewrite, or a return to an earlier version. Each edit is located in the current text and reported as found, not found, ambiguous or overlapping. An edit can name the comment it answers, and that comment gets a reply |
withdraw_proposal | Takes back a proposal you have not reviewed yet |
add_comments | Adds comments, questions, issues or highlights to specific passages, as one group attributed to the app |
reply_to_comment | Replies inside an existing comment thread, optionally resolving it |
submit_step | Submits one finished step of a running method, which MindMax validates and anchors before saving |
Write
| Tool | What it does |
|---|---|
create_document | Creates a new document, optionally at an exact path. Refuses an existing path instead of replacing it |
append_to_document | Appends to the end of a note, the end of one of its sections, or today's daily note |
move_document | Renames or moves a file or folder through MindMax, so history, comments and open tabs follow it, with an undo |
create_canvas | Turns a Markdown outline into a canvas |
add_canvas_cards | Adds cards to a canvas, at the top level or growing from a card, signed with the app's name |
revise_canvas_card | Revises a card the app wrote itself; on MindMax's cards it may only add points |
link_canvas_cards | Draws support, contradiction or reference relations between existing cards |
grade_canvas_answers | Marks answers you wrote on question cards, with a verdict and a note |
export_document | Exports through MindMax's own exporter — Word, PDF or a self-contained HTML page — beside the original |
add_bookmark | Bookmarks a document or a saved search, optionally into a named group |
start_pipeline | Starts one of MindMax's methods: MindMax plans the steps and validates the results, the app's model does the thinking |
Methods an app can run with its own model
These are the jobs MindMax normally runs with its own model. On this path the connected app's model does the work step by step; MindMax supplies the instructions, the chunking and the same validators, and saves the result where its own run would have saved it.
| Method | What it produces |
|---|---|
reading_guide | A full annotation pass with a chosen lens — guide, review, critique or study — with margin notes and a closing write-up, in MindMax's reading panel |
comment_answers | An answer in each comment thread that was waiting for one |
comment_insights | An insights note distilled from your highlights and comments, with replies on the threads it drew from |
doc_map_notes | Chapter and node notes for a long book, written into the map MindMax uses for its own answers |
canvas_skill | One thinking method applied onto a canvas, in the method's own shape |
translation | A whole-document translation with one terminology list, in the same side-by-side view as a MindMax run |
ocr | Recognised text for scanned pages, page by page, with the same per-page quality record |
A method runs as steps. The app asks for a step, does the work, submits it, and gets the next ones; steps marked parallel can go to sub-agents. An output that fails validation comes back with the reason and can be resubmitted, twice.
Prompts
Hosts that support prompts — Claude lists them under the connector — offer these as menu items:
| Prompt | What it does |
|---|---|
proofread | Proofreads a document and sends the fixes as suggestions |
writing-review | Reads a draft as an editor, a peer and a manager would, and leaves the problems as comments without changing the text |
resolve-comments | Answers the comments waiting on a document, checks their facts, and proposes the edits they ask for |
verify-claims | Checks a document's claims against your library and leaves the verdicts as comments |
write-from-library | Researches a topic in your library and writes a new, cited document |
comment-insights | Distils your highlights and comments on a document into an insights note |
prepare-book | Writes a long book's chapter map, then a reading guide with priorities and margin notes |
How a call is shaped
- Workspace. Every tool takes an optional workspace, by name or folder. Omitted, it means the workspace you have in front — or the most recent one, if MindMax is in the background.
- Paths. Absolute and workspace-relative paths both work. Writes must stay inside the workspace.
- Quote handles. A search hit comes back with a handle such as
q7. An app cites it as【path·heading||q7】, and MindMax expands that into the verbatim passage with a clickable anchor when the text is saved. A handle that was never issued is dropped rather than written into your document. - Limits. A single result is capped so one tool call cannot fill the app's context; the tail of the result says how to continue.
- No models of ours. Nothing here calls a MindMax model, so nothing here spends credits. Where a job needs a model, the app's own model does it.