add_cards_to_inbox
ChatGPTPut cards in the user's Heptabase Inbox to process later, like the app's Add to Inbox. A card already in the Inbox moves back to the top, out of Done. With remindDate the cards wait in Later until that date. Cards you create are not added to the Inbox; call this when the user wants them there. Before first use, read the inbox page with read_system_manual_page.
align_whiteboard_objects
ChatGPTAlign target whiteboard objects by an edge or center. Optional reference objects stay stationary and contribute their combined bounds; without references, targets align within their own selection bounds. Requires an open Heptabase client.
append_to_journal
ChatGPTAppend content to a journal in Heptabase. If the journal for the specified date does not exist, it will be created. IMPORTANT: Pass the date parameter with the user's current local date (YYYY-MM-DD) to avoid timezone mismatch.
arrange_whiteboard_objects
ChatGPTArrange existing whiteboard objects into an ordered row, column, or grid. Connections cannot be arranged directly; arrange their endpoint objects instead. Requires an open Heptabase client.
create_audio_card_from_file
ChatGPTCreate an audio card in the user's Heptabase space from an audio file imported with import_file_from_upload or import_file_from_url. A file's existing audio card is reused, or restored if it is in the trash, so retrying never creates a duplicate. It does not transcribe the audio.
create_connections
ChatGPTCreate up to 100 labeled or unlabeled connections between pairs of objects on a whiteboard. Requires an open Heptabase client.
create_database
ChatGPTCreate a new tag database: a tag with one table view. Fails when a tag with the same name already exists in the space.
create_image_card_from_file
ChatGPTCreate an image card in the user's Heptabase space from an image imported with import_file_from_upload or import_file_from_url. A file's existing image card is reused, or restored if it is in the trash, so retrying never creates a duplicate.
create_mind_map
ChatGPTCreate a complete mind map on a whiteboard from text, regular Cards, highlights, and text elements. existingCard and cardInstance accept only regular Cards reported as card, not pdfCard or other source or media card types. For content read from an unsupported card type, use textNode by default, or newCard when the result should be an independently editable Card. For cardInstance, highlightElementInstance, or textElement, copy the canonical ID or exact inst:<placementId> from whiteboard read output. Requires an open Heptabase client.
create_object
ChatGPTCreate a new note card or fill a missing or empty daily journal in the user's Heptabase space. Existing journal content is never overwritten. Write content in Hepta Markdown.
create_pdf_card_from_file
ChatGPTCreate a PDF card in the user's Heptabase space from a PDF imported with import_file_from_upload or import_file_from_url. A file's existing PDF card is reused, or restored if it is in the trash, so retrying never creates a duplicate. It does not parse the PDF; call parse_pdf to read its pages.
create_section
ChatGPTCreate a new section around up to 100 objects already on a whiteboard. The new section also includes any other objects fully inside its bounds. Requires an open Heptabase client.
create_skill
ChatGPTCreate a Heptabase Skill: a repeatable workflow stored as an instruction card in the user's own main space, which the Heptabase AI Agent and every connected agent can run. Use it only when the user asks to save something as a skill and either mentions Heptabase or the workflow handles content in their knowledge base; otherwise use your host's own skills. Never suggest creating a Heptabase Skill on your own. Instructions are Hepta Markdown. Fails when the command name is taken by an available Skill or reserved for a System Skill. Before first use, read the skills page with read_system_manual_page.
create_tag_group
ChatGPTCreate a tag group, a top-level section in the Tags sidebar for organizing tag trees. A tag group is not a tag and cannot be used as a parent tag.
create_video_card_from_file
ChatGPTCreate a video card in the user's Heptabase space from a video imported with import_file_from_upload or import_file_from_url. A file's existing video card is reused, or restored if it is in the trash, so retrying never creates a duplicate. It does not transcribe the video.
create_video_card_from_url
ChatGPTSave a YouTube video as a video card in the user's Heptabase space. An existing video card for the same URL is reused, or restored if it is in the trash, so retrying never creates a duplicate. Heptabase saves the video captions as its transcript in the background; read it with read_object.
create_web_card_from_url
ChatGPTSave a web page as a web card in the user's Heptabase space. An existing web card for the same URL is reused, or restored if it is in the trash, so retrying never creates a duplicate. Heptabase extracts the page content in the background when the card has none; read it with read_object, or open the URL yourself if you need the page now.
create_whiteboard
ChatGPTCreate a root whiteboard, or a sub-whiteboard when parentWhiteboardId is given. Requires an open Heptabase client.
create_whiteboard_shortcut
ChatGPTShow an existing whiteboard on another whiteboard as a linked shortcut, without changing its hierarchy placement. Requires an open Heptabase client.
delete_tags
ChatGPTPermanently delete tags, each together with all its nested tags. Every deleted tag also loses its database, that database's views, and the properties only that database uses along with their values; cards stay but lose the tag. This cannot be undone, and Revert does not restore deleted tags. Use it only when the user explicitly names the tags to delete. Before calling, tell the user which tags and nested tags will go and how many cards carry them, say that their databases and property values will be lost for good, and wait for the user to confirm. Requires an open Heptabase client.
edit_card_properties
ChatGPTEdit property values on cards that already belong to a Heptabase tag database. Use read_database first to get the database identifiers, property IDs and types, select options, and relation target database IDs. Edits are evaluated independently and in input order. Use null to clear a property.
edit_database
ChatGPTEdit the configuration of an existing tag database: its name, properties, select options, views, filters, sorts, and kanban grouping, or attach an existing shared property. Read the database first for its current property, option, and view IDs. Edits apply in order; to create a property, view, or select option, give it a CREATE- placeholder ID such as CREATE-status and reuse that placeholder wherever later edits in the same call refer to it. Requires an open Heptabase client.
edit_journal_content
ChatGPTEdit a journal's content in Hepta Markdown. Put every change to the journal in one call: each replace is matched against the content as read, and the call is written whole or not at all, reporting every failed edit. A journal that does not exist yet is created with the edited content.
edit_note_card_content
ChatGPTEdit a note card's content in Hepta Markdown. Put every change to the card in one call: each replace is matched against the content as read, and the call is written whole or not at all, reporting every failed edit.
edit_object_content
ChatGPTReplace exact content in a note card or journal. Read the object first, remove only line-number prefixes, and use the shortest unique oldString. The tool preserves supported formatting and returns an excerpt of the applied edit when available.
get_file_download_url
ChatGPTGet a temporary download URL for an original file stored in Heptabase. Pass a fileId from a read, list, or search result and its containing source object. Download the bytes yourself using HTTP; this tool does not return file contents. Supports uploaded PDFs, videos, audio, images, and attachments. Request a new URL when it expires.
get_file_upload_url
ChatGPTGet a temporary URL to upload a local file to Heptabase. Use it only when you can run shell commands; otherwise ask the user to add the file in Heptabase. PUT the exact bytes to uploadUrl, then call import_file_from_upload.
import_file_from_upload
ChatGPTTurn a file uploaded with get_file_upload_url into a Heptabase file. Any file type can be imported; fileKind says which card tool can use it. Importing the same fileId again returns the same file.
import_file_from_url
ChatGPTDownload a file from a public http or https URL, up to 50 MB, and turn it into a Heptabase file. Any file type can be imported; fileKind says which card tool can use it. Each call creates a new file, so do not retry a call that succeeded. For a larger file or one that needs a login, upload it with get_file_upload_url instead.
lint_whiteboard
ChatGPTCheck a whiteboard for deterministic object overlap, section-boundary, connection-crossing, and missing visible connection-shaft problems. Requires an open Heptabase client.
list_cards
ChatGPTList cards by type, whiteboard or tag membership, timestamps, or known IDs without reading their content. When more cards match than the current page, the result includes the total matching count.
list_inbox_items
ChatGPTList items in the user's Heptabase Inbox: cards and messages the user added, mentions, replies, quotes, and invites. Returns compact lines with each item's inboxItemId; read a card's content with read_object. Before first use, read the inbox page with read_system_manual_page.
list_skills
ChatGPTList the Heptabase Skills available in the user's main space, with each Skill's slash command, when-to-use description, skill ID, and instruction card ID. The list also comes with Space Instructions, so use this when that list says more are available or to page through them.
list_tags
ChatGPTList all tags in the user's Heptabase knowledge base with card counts and hierarchy. USE WHEN: - You need to discover what tags exist before drilling into a specific tag - User asks "what are my tags?" or "how is my knowledge organized?" - You want to find the right tag ID before calling list_cards RETURNS: - All tags grouped by tag group (if any) - Card count per tag - Parent-child tag hierarchy - Tag IDs for use with list_cards or read_database STRATEGY: - Call this first to discover tags, then use list_cards with tagIds to list a specific tag's cards - Use read_database when you need a tag database's property schema or views - Use the optional nameFilter to narrow results if the user mentions a topic
list_whiteboards
ChatGPTList the user's whiteboard hierarchy with lightweight object counts and pagination.
move_tags
ChatGPTMove one or more existing tag trees under a parent tag, into a tag group, or to the ungrouped root. All selected roots share one destination, and their descendants remain attached. Moving under a parent automatically adopts that parent's tag group. Moving to a tag group or ungrouped makes each selected tag a root. Find tag and tag group IDs with list_tags.
move_whiteboard
ChatGPTMove an existing whiteboard in the whiteboard hierarchy. This moves the hierarchy placement, not linked whiteboard shortcuts. Requires an open Heptabase client.
move_whiteboard_objects
ChatGPTMove existing objects within a whiteboard while preserving each group's internal layout. Connections cannot move directly; move their endpoint objects instead. Requires an open Heptabase client.
move_whiteboard_objects_across_whiteboards
ChatGPTMove objects from one whiteboard to another while preserving their relative layout. A selected section includes all descendants. Connections and section memberships are preserved when both related objects move; relations crossing the selection boundary are removed. The group is appended to the right of existing destination content. Requires an open Heptabase client.
navigate_heptabase
ChatGPTChange what the Heptabase app that is open in this chat shows, without opening a new copy: another whiteboard, or a card in the card panel beside the whiteboard. Pass the view ID from the latest open_heptabase result in this chat. Use it when the user asks to see something, or to show a whiteboard that you are about to build or change, such as one you just created. If it fails with viewNotOpen, call open_heptabase. The user can also switch whiteboards and open cards in the app at any time, so its result may be out of date: before you later read, change, or answer about a whiteboard or card that the user does not name, call read_heptabase_view first.
open_heptabase
ChatGPTOpen Heptabase for the user inside the chat. It shows one editable whiteboard; the user can read, arrange and edit it, and add items from it to the chat. Use this tool when the user wants to see, browse, edit, or arrange their notes themselves, or to let the user watch you build or change a whiteboard. Once it has loaded, it runs the tools that require an open Heptabase client, such as the whiteboard editing tools, so open it before them when no Heptabase app is open. To read or change content for the user, use the other tools. Each call opens a new copy of Heptabase in the chat, which takes a few seconds to load. If Heptabase is already open in this chat, use navigate_heptabase with its view ID instead, unless the user asks for a new copy or navigate_heptabase fails with viewNotOpen. The user can switch whiteboards and open cards in it at any time, and you are not told, so before you read, change, or answer about a whiteboard or card that the user does not name, call read_heptabase_view with its view ID.
parse_pdf
ChatGPTParse a PDF card so read_pdf can read its pages. Parsing runs in the background and can take a few minutes for long PDFs. While it runs, call parse_pdf again to check progress; that call also finishes the parse when it is ready. A parsed PDF is not parsed again.
place_whiteboard_objects
ChatGPTAutomatically place up to 100 existing objects on a whiteboard. Split larger batches across calls. Accepted objectType values are card, journal, pdfCard, imageCard, videoCard, audioCard, and webCard. The batch is packed left-to-right immediately to the right of persisted content, and every valid item creates a new instance even when it is already present. This auto-only phase does not infer Section membership or adjust Sections, neighboring objects, connections, or lint. Retrying may create duplicates.
query_database
ChatGPTQuery rows from a tag database with optional filters, sorts, and paging. Read the database first for its property IDs, select option IDs, and relation fields. Provide the database ID or tag ID, and list the fields to return. Requires an open Heptabase client.
read_backlinks
ChatGPTFind the backlinks to a card, journal, or whiteboard in the user's Heptabase knowledge base: the cards, journals, and whiteboards that link to or mention it, or that place, group, or connect a card on a whiteboard. Each backlink is an excerpt; use read_object or read_whiteboard to read a source in full. Pages by backlink group: pass the returned next offset to read more. Requires an open Heptabase client.
read_card_highlights
ChatGPTRead the highlights on a note card or PDF card in the user's Heptabase knowledge base, with each highlight's note. Use offset and limit (at most 50) to page through them.
read_database
ChatGPTRead a tag database configuration by database ID, tag ID, or both, including its property schema, views, filters, sorts, select options, and relation targets.
read_heptabase_view
ChatGPTSee what the Heptabase app open in this chat shows right now: the whiteboard on screen, and the card open in the card panel beside it, if any. The user can switch whiteboards and open cards in the app at any time, and you are not told, so earlier open_heptabase and navigate_heptabase results may be out of date. Always call this first before you read, change, or answer about a whiteboard or card that the user does not name, such as "the whiteboard", "this whiteboard", "this card", "here", or "what I'm looking at". Then use the whiteboard or card it returns, even if it is not the one you opened. Pass the view ID from the latest open_heptabase result in this chat.
read_image
ChatGPTRead an image referenced by read_object. Pass the file and source object fields from the image reference.
read_journal_range
ChatGPTRead existing daily journal entries within an inclusive date range from the user's Heptabase knowledge base, oldest first. USE WHEN: - User asks about their journal entries during a time period - User wants to see what they wrote in past days, weeks, or months - User needs to review their daily notes IMPORTANT CONSTRAINTS: - Each call can retrieve at most 92 days (approximately 3 months) - For longer periods, make multiple calls - Dates use YYYY-MM-DD format. Both startDate and endDate are inclusive.
read_object
ChatGPTRead a specific object from the user's Heptabase knowledge base. Use this after another tool returns an object ID. Text content uses 1-based absolute line numbers for reference; the number prefixes are not part of the content. Use offset and limit to continue reading long content. For a PDF card, this returns metadata; use read_pdf to read its pages.
read_pdf
ChatGPTRead specific pages from a parsed PDF card in the user's Heptabase knowledge base. Page numbers are 1-indexed and inclusive.
read_space_instructions
ChatGPTRead the space's Space Instructions card: the user's standing rules for how agents work in this space. Read it once before you first use other Heptabase tools in a conversation, and follow it over the system manual's defaults. To change the rules, edit the returned card with edit_object_content. Text content uses 1-based line numbers for reference; the number prefixes are not part of the content.
read_system_manual_page
ChatGPTRead Heptabase reference documentation. concepts explains cards, whiteboards, and tag databases. hepta_markdown defines the rich-text syntax for tables, object mentions, and text or background colors used by create_object, append_to_journal, and edit_object_content. properties documents database property types and editable values. Omit path to list all three pages.
read_whiteboard
ChatGPTRead a whiteboard's structure or object content from the user's Heptabase knowledge base. Use structure mode to inspect its organization, or content mode to read and analyze the objects on it.
read_whiteboard_layout
ChatGPTRead the spatial layout of a whiteboard. Returns each object with its title, ID, position, and size. Does not return object content; use read_whiteboard for content. Requires an open Heptabase client.
remove_inbox_items
ChatGPTRemove items the user added to the Heptabase Inbox, like the app's Remove from Inbox, by the inboxItemIds from list_inbox_items. The cards and messages themselves stay. Mentions, replies, quotes, and invites cannot be removed; mark them done with update_inbox_items instead. When you chose which items to remove, list them and wait for the user to confirm first. Before first use, read the inbox page with read_system_manual_page.
remove_whiteboard_objects
ChatGPTRemove object placements from a whiteboard without deleting their underlying cards. For sections, the members are kept (only the section frame is removed). Connections attached to removed objects and section memberships are cleaned up automatically. Only linked whiteboard shortcuts can be removed; embedded sub-whiteboards cannot. Requires an open Heptabase client.
rename_object
ChatGPTRename a whiteboard, section, or source card (PDF, web, image, video, or audio). Note card and journal titles derive from the first heading in their content, so use edit_object_content instead. Requires an open Heptabase client.
render_cards
ChatGPTPresent selected Heptabase cards in an interactive UI. After searching, recommending, creating, or editing cards, render the card or small set selected for the user before answering. Complete requested changes first. Use in clients that support MCP Apps; in plain-text/CLI clients, respond in text instead. Supports notes, journals, PDF, video, audio, image, web cards, and highlights.
resize_whiteboard_objects
ChatGPTResize, fit to content, reset to default size, or fold or expand objects already on a whiteboard. Mind maps and connections cannot be resized. Requires an open Heptabase client.
restore_cards
ChatGPTRestore one or more cards from the trash. Their whiteboard placements and mind map nodes that were not removed show them again.
restore_whiteboards
ChatGPTRestore one or more whiteboards from the trash, with the nested whiteboards that went to the trash with them. A whiteboard nested in a whiteboard that stays in the trash moves to the top level. Requires an open Heptabase client.
screenshot_whiteboard
ChatGPTTake a screenshot of a whiteboard, simplified for an agent to check layout. Unlike the app view, each object is drawn as its bounds with its title and type, together with sections, connection routes, and connection labels; object content and the internal structure of mind maps are not drawn. Use it after deliberate layout changes to check reading flow, grouping, alignment, whitespace, section hierarchy, connection routing, and color balance. Requires an open Heptabase client.
search_by_keyword
ChatGPTSearch the user's entire main Heptabase Space by exact keywords or short phrases. It searches titles and supported text content, while PDFs, video cards, and audio cards are title-only. Results include concise object summaries and matching excerpts; use read_object or read_pdf for complete content.
search_by_semantic
ChatGPTSearch the user's Heptabase space semantically using a natural-language query. Use this for ideas, concepts, or meanings that exact keywords may miss. Results contain concise object summaries and excerpts; use read_object or read_pdf when complete content is needed. This does not search tags or whiteboards.
search_object_content
ChatGPTSearch literal or regex text inside one known object in the user's Heptabase knowledge base. Returns excerpts with absolute 1-based line numbers, marks matching lines with *, and searches stored parsed page content for PDF cards. Use another search tool or read_object first to obtain the object ID and type.
set_space_instructions
ChatGPTMake an existing note card this space's Space Instructions: the user's standing rules for how agents work in this space. If another card is the Space Instructions, it stops being one; that card itself does not change. When the user asks to make an existing card the Space Instructions, set that card. Otherwise, when the space has no Space Instructions and the user asks, agrees, or states a lasting rule for this space, create the note card with create_object, then set it. When the space already has Space Instructions, add rules by editing that card with edit_object_content instead. Before first use, read the space_instructions page with read_system_manual_page.
start_web_app_session
ChatGPTUsed by the Heptabase app to start its restricted-session sign-in. It is not useful to call directly.
trash_cards
ChatGPTMove one or more cards to the trash. The cards keep their whiteboard placements and mind map nodes, so restoring a card brings it back where it was; the result lists those placements. Journals cannot be moved to the trash. When you chose which of the user's existing cards to trash, for example duplicates to clean up, list them and wait for the user to confirm first. After trashing, remove the listed whiteboard placements with remove_whiteboard_objects unless the user wants them kept.
trash_whiteboards
ChatGPTMove one or more whiteboards to the trash. Each whiteboard takes the whiteboards nested in it along, and restoring it with restore_whiteboards brings them back; the result lists them. Cards, sections, mind maps and other objects on these whiteboards stay, and the cards stay in the card library. When you chose which of the user's existing whiteboards to trash, for example empty ones to clean up, list each one with the nested whiteboards that would go with it and wait for the user to confirm first; a whiteboard with nested whiteboards is not empty. After trashing, tell the user which nested whiteboards went along. Requires an open Heptabase client.
update_connection
ChatGPTUpdate the route or color of one existing whiteboard connection while preserving its endpoints and other properties. Requires an open Heptabase client.
update_database_card_membership
ChatGPTAdd up to 200 cards to an existing Heptabase tag database. Use list_tags or read_database to get the database ID or tag ID. Use card listing, search, read, or create results to get card IDs. Repeated additions are skipped. Remove is not yet supported.
update_inbox_items
ChatGPTMark items in the user's Heptabase Inbox done, read, or unread, remind later, or clear a remind date, by the inboxItemIds from list_inbox_items. Marking a card done does not file it; place it on a whiteboard or tag it first. When you chose which items to change, list them and wait for the user to confirm first. Before first use, read the inbox page with read_system_manual_page.
update_mind_map
ChatGPTEdit the structure of one existing mind map in a single atomic change: add, relabel, move, or delete nodes, and change its layout, collapsed branches, or edge colors. Operations apply in order, and nothing changes unless every operation succeeds. Read the mind map first and target existing structure by mindMapNodeId. A later operation may reference a node added earlier in the same call by its nodeKey. updateTextNode fully replaces only an intrinsic textNode plain-text label; use edit_object_content for Card content. This tool does not edit Text Element or Highlight content, replace the whole tree, or convert node types. Card nodes support only regular Cards reported as card, not PDF or other source or media Cards. For a consumed Card Instance, Highlight Element Instance, or Text Element, copy the canonical ID or exact inst:<placementId> from whiteboard read output. Requires an open Heptabase client.
update_skill
ChatGPTChange a Heptabase Skill's command name or when-to-use description. Only the Skill's creator can change it. To change its instructions or display name, edit the content of its instruction card instead: the display name is the card's first-line title, # Skill: Name. Before first use, read the skills page with read_system_manual_page.
update_whiteboard_objects_color
ChatGPTUpdate only the color of objects on a whiteboard. Connections are supported; highlight elements and mind maps are not. Requires an open Heptabase client.
add_cards_to_inbox
Claudealign_whiteboard_objects
Claudeappend_to_journal
Claudearrange_whiteboard_objects
Claudecreate_audio_card_from_file
Claudecreate_connections
Claudecreate_database
Claudecreate_image_card_from_file
Claudecreate_mind_map
Claudecreate_object
Claudecreate_pdf_card_from_file
Claudecreate_section
Claudecreate_skill
Claudecreate_tag_group
Claudecreate_video_card_from_file
Claudecreate_video_card_from_url
Claudecreate_web_card_from_url
Claudecreate_whiteboard
Claudecreate_whiteboard_shortcut
Claudedelete_tags
Claudeedit_card_properties
Claudeedit_database
Claudeedit_journal_content
Claudeedit_note_card_content
Claudeedit_object_content
Claudeget_file_download_url
Claudeget_file_upload_url
Claudeimport_file_from_upload
Claudeimport_file_from_url
Claudelint_whiteboard
Claudelist_cards
Claudelist_inbox_items
Claudelist_skills
Claudelist_tags
Claudelist_whiteboards
Claudemove_tags
Claudemove_whiteboard
Claudemove_whiteboard_objects
Claudemove_whiteboard_objects_across_whiteboards
Claudenavigate_heptabase
Claudeopen_heptabase
Claudeparse_pdf
Claudeplace_whiteboard_objects
Claudequery_database
Clauderead_backlinks
Clauderead_card_highlights
Clauderead_database
Clauderead_heptabase_view
Clauderead_image
Clauderead_journal_range
Clauderead_object
Clauderead_pdf
Clauderead_space_instructions
Clauderead_system_manual_page
Clauderead_whiteboard
Clauderead_whiteboard_layout
Clauderemove_inbox_items
Clauderemove_whiteboard_objects
Clauderename_object
Clauderender_cards
Clauderesize_whiteboard_objects
Clauderestore_cards
Clauderestore_whiteboards
Claudescreenshot_whiteboard
Claudesearch_by_keyword
Claudesearch_by_semantic
Claudesearch_object_content
Claudeset_space_instructions
Claudetrash_cards
Claudetrash_whiteboards
Claudeupdate_connection
Claudeupdate_database_card_membership
Claudeupdate_inbox_items
Claudeupdate_mind_map
Claudeupdate_skill
Claudeupdate_whiteboard_objects_color
Claude