create_board
ChatGPTCreate a board (visual canvas), optionally with its full initial content in this single call: cards with blocks, loose blocks, sub-boards, and connections between them. Layout is computed automatically from that structure, so no coordinates are needed. layout picks the arrangement: grid (default, reading order), mindmap (hub and branches from connections, optional root card index), columns (cards grouped by their column field), moodboard (masonry packing, suited to images). Creates new boards only; edit_board changes existing ones.
create_deck
ChatGPTCreate a presentation (16:9 slides) from layout slides: each slide is a layout (title, section, title_body, bullets, columns, image_text, image_full, quote, stats, big_number, process, timeline, comparison, chart, compose) plus its slots, and Draftt places the content. One theme (colors, fonts) applies to the whole deck. The response lists warnings per slide (text too long for its slide, low contrast). For when the user asks for a new presentation.
create_doc
ChatGPTCreate a doc (a long-form document: briefs, scripts, notes, reports) inside a board. The board shows it as a doc tile; people open it in a side panel or as its own page. The body is HTML or plain text. For when the user asks for a new doc; edit_doc changes existing ones.
create_upload_urls
ChatGPTPrepare uploads of local files (photos, PDFs, audio, video) to Draftt, for clients that can run shell commands. Returns, per file, an upload_url and the url the file will have. A file goes up with: curl -X PUT -H "Content-Type: <content_type>" --data-binary @<path> "<upload_url>". The content type and the exact byte size must match what was declared, and the upload URLs expire after 10 minutes. The returned url works wherever an image or file goes: image_url and image elements in create_deck and edit_deck, or the url of an add_block op in edit_board. Works for boards and presentations; up to 50 files and 25MB each. A file already online needs no upload: its public URL works directly.
edit_board
ChatGPTEdit an existing board with a batch of typed operations in one call: add_card, add_block, update_card, update_block, update_checklist_item (one to-do: text, done, assignee, due), connect, arrange, or set_frame (coordinates and dimensions; cards accept x, y and width, loose blocks also height). Ops apply sequentially with no rollback. Nodes are referenced by UUID, the 8-char short id from get_board, or a ref declared earlier in the same call. upload_board_asset stores a binary file first; create_board makes new boards.
edit_deck
ChatGPTEdit a presentation with a batch of ops, and optionally rename it (name). add_slide / replace_slide take a layout and its slots, and Draftt places everything on the 1920x1080 slide. Also delete_slide, move_slide, set_background, set_hidden (a hidden slide stays in the deck but is skipped when presenting and on public links), and for fine adjustments add_element / update_element / delete_element (boxes in slide px; a line shape can carry arrowheads; type chart adds a real, editable chart: column, bar, line, area, pie or donut). get_deck gives the slide ids. The response lists warnings for every slide touched (overflow, overlaps, text off the slide).
edit_doc
ChatGPTEdit a doc: rename it (name), replace its whole body (content) or add to the end (append). Changing part of the body means sending the full HTML with the edits; get_doc returns the current one. For when the user asks to change a doc.
get_board
ChatGPTRead one board as a semantic outline plus a compact visual_layout scene graph. visual_layout includes canvas bounds and every card or loose-block frame, aspect ratio, and short ref. Block references look like [text 1a2b3c4d]; sub-board lines include the full board_id. File and image URLs are included. Carries updated_at, last_edited_by, last_viewed_by_you and the current time (now). get_board_snapshot renders the same board as an image; edit_board and upload_board_asset change it.
get_board_snapshot
ChatGPTRender a board or selected canvas nodes as a PNG image, to inspect the actual composition. get_board gives the structure, geometry and node ids. Without node_ids it shows an overview; a smaller set of nodes shows more detail.
get_deck
ChatGPTRead a presentation: its slides in order, each with its background, its elements in paint order (id, short_id, type, box x/y/w/h on the 1920x1080 slide, text, font size and color, shape fill, image url) and warnings (text running off the slide, overlapping text, text under 16px, low contrast). Slides hidden from presenting carry hidden: true (edit_deck set_hidden changes it). Slides with unresolved comment threads carry open_comments (get_deck_comments has them). The deck and each slide carry updated_at, the deck last_edited_by and last_viewed_by_you, and now is the current time. Returns the slide ids, colors and fonts edit_deck works with.
get_deck_snapshot
ChatGPTRender slides of a presentation as PNG images (960x540 each, up to 8 per call) with the deck's fonts, showing what the audience sees: wraps, balance, crowding, contrast.
get_doc
ChatGPTRead a doc (a long-form document): its title (name) and its body as HTML (p, h1-h3, ul/ol/li, blockquote, hr, strong, em, a, img, details toggles). Docs live inside boards as sub-boards: get_board lists them. Carries updated_at, last_edited_by, last_viewed_by_you and now.
list_boards
ChatGPTList the boards (visual canvases with cards and blocks) and presentations (kind "deck") the user can open, across every workspace they are in (shared boards included), most recently updated first. Slim: { id, name, kind, folder_id, folder_name, card_count, block_count, updated_at, last_edited_by, last_viewed_by_you, workspace_id, workspace_name } per item, plus the current time in now. get_board and get_deck read each one. Sub-boards are not listed here; they appear inside their parent board via get_board. No cursor: workspace_id or name_query narrow the list when has_more is true.
list_folders
ChatGPTList the folders of the user's teams. Folders are optional and only sort a team's boards and presentations; the whole team sees what is in the team, at its root or in a folder. A folder id works in create_board / create_deck to create there, and in list_boards to see what is in it.
list_tasks
ChatGPTList the open checklist items that are the user's own: the ones assigned to them, and the ones with a day they set themselves with nobody assigned, across every board they can open. Soonest day first, tasks with no day last; overdue is true when the day has passed in the user's calendar. Shows up to 50 tasks; has_more says there are more (due_before or workspace_id narrow it). For when the user asks what they have pending. Each task carries the board and block ids and the short ids of its block and item, which edit_board's update_checklist_item takes.
list_workspaces
ChatGPTList the workspaces the user belongs to, most recently used first, with their role and whether new boards can be created there. An id works as workspace_id in list_boards (to narrow the list) and in create_board / create_deck (to choose where it goes; by default it goes to the first one with can_create).
search_catalog
ChatGPTSearch Explore, Draftt's curated catalog of frames from published commercials, music videos and films, each with its credits, by what the frames show, a work, brand or credited person, a HEX color, or likeness to another frame (similar_to). The user sees the results as a visual grid. For when the user wants references or a moodboard for an idea; each search covers one idea (the subject, the format, the look). An item's url works in an edit_board image block or on slides, where the frame stays linked to the catalog, never copied.
search_library
ChatGPTSearch the user's library in Draftt (Guardados): the frames they saved from Explore and the images they or their team uploaded or imported, by what the images show, a HEX color, or likeness to another image (similar_to). The user sees the results as a visual grid. For when the user asks for their saves, uploads or references ("my red references", "what did I save for the car spot"). An item's url works in an edit_board image block or on slides. Only images are searched; the library holds no videos or PDFs here.
upload_board_asset
ChatGPTDownload a file from a public URL into the board's own storage, so it survives even if the source disappears. Returns the stored asset's url, which goes in an add_block op of edit_board (type image, file, or audio). Limits: 25MB max; media and common document types only (no HTML); private/internal hosts are blocked. A link that should stay a link (a YouTube video, a website) needs no upload: it goes in an add_block op of edit_board (type embed).