MCP App Store
by KeenEthics
Zoom Canvas icon

Zoom Canvas MCP Server for Claude and ChatGPT

By Zoom

Overview

Connect Zoom Canvas to Claude to create, find, edit, and share documents without leaving your conversation. Turn ideas into drafts, refine content based on comments, and preview updates right in the chat. Claude can also manage data tables, organize files and pages, add collaborators, adjust sharing permissions, and transfer ownership. From updating project plans to sharing team content, just describe what you need and keep work moving.

How to connect

  1. Claude: Open the Zoom Canvas listing in the Claude connectors directory and choose Connect. Claude listing Zoom Canvas exposes 34 tools here.
  2. ChatGPT: Open the Zoom Canvas listing in the ChatGPT app directory and choose Connect. ChatGPT listing Zoom Canvas exposes 34 tools here.

Tools

canvas_add_file_collaborators

ChatGPT
Tool Description Adds collaborators (Zoom users, Team Chat channels, or email addresses) to a Zoom Canvas (formerly Zoom Docs) file, granting them a specified access role. Only doc or data table file types are supported. When to Use Use this tool when a user wants to share a Canvas file with one or more people or channels by assigning them viewer, commenter, editor, or co-owner permissions. Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. You MUST follow this exact sequence: - Step 1 (ask): Present the target file (show key identifying info such as file name or file ID) and the full list of collaborators to be added, including each collaborator's identifier (user_id / channel_id / email) and the role being granted. Clearly state that this will add these collaborators with the specified access, and ask "Do you want to proceed?". Do NOT call the add-collaborators tool in this step. - Step 2 (execute): Only call the tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "add them"). If the user does not confirm, says no, or asks to change any collaborator/role, abort immediately and do not call the tool. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. - If the user changes the collaborator list or roles after Step 1 (e.g., adds/removes a person, changes a role), treat this as a new request: return to Step 1 and present the updated list for fresh confirmation before calling the tool. Must Do - Each collaborator entry must include exactly one identifier: user_id, channel_id, or email - Each collaborator entry must include a role (viewer, commenter, editor, or co-owner) - Provide between 1 and 50 collaborators per request - The target file must be of type doc or data table Must NOT Do - Do not include more than one identifier field (user_id, channel_id, email) per collaborator entry - Do not use this tool on unsupported file types (e.g., spreadsheets, presentations)

canvas_create_comment

ChatGPT
Tool Description Creates a root comment or a reply in a specified Zoom Canvas file, using plain text content. When to Use Use this tool when a user wants to: - create a new root comment thread (discussion or full_page), or - reply to an existing thread, or - reply to a specific existing comment. Prerequisites - Must have Commenter or higher access to the target Canvas file. - The fileId of the target Zoom Canvas file must be known. Must Do - Provide exactly one of thread_type or parent per request — never both. - Set content.format to plain_text. - Keep comment text to 40,000 characters or fewer. Reply Target Rules (STRICT) When creating a root comment: - provide thread_type - do not provide parent When replying to a thread: - provide parent.type = "thread" - provide parent.thread_id - do not provide parent.parent_comment_id - do not provide thread_type When replying to a specific comment: - provide parent.type = "comment" - provide parent.parent_comment_id - do not provide parent.thread_id - do not provide thread_type Mixed combinations are invalid. Semantics - Replying to a thread attaches the new comment to the thread. - Replying to a specific comment attaches the new comment directly under that comment. - If the user asks to "reply to comment X", prefer replying to the specific comment rather than the thread containing it. Must NOT Do - Do not use thread_type when creating a reply. - Do not attempt to create inline root threads — this endpoint only supports root discussion and full_page threads. - Do not send both thread_type and parent. - Do not mix parent.thread_id and parent.parent_comment_id. - Do not send parent.type="thread" with parent.parent_comment_id. - Do not send parent.type="comment" with parent.thread_id. Valid Request Shapes | Intent | Required fields | Forbidden fields | |---|---|---| | Create root discussion comment | thread_type="discussion", content | parent | | Create root full-page comment | thread_type="full_page", content | parent | | Reply to thread | parent.type="thread", parent.thread_id, content | thread_type, parent.parent_comment_id | | Reply to comment | parent.type="comment", parent.parent_comment_id, content | thread_type, parent.thread_id |

canvas_create_file

ChatGPT
Create a new file, include a doc, a folder, or a data table; for data table, you can specify the columns to be created. Returns the new file's file_id which is required for all subsequent operations on the file. When to Use - The user wants to create a brand-new data table file from scratch. - The user wants to create a new doc, folder. When NOT to Use - Do not use this tool to create a table inside an existing file. Use canvas_data_table_create_data_table instead. Tool-Specific Rules - To start a full data table workflow, first call canvas_create_file with file_type = "data_table", then use canvas_data_table_create_data_table to add tables (optional), canvas_data_table_create_column to define columns, and canvas_data_table_create_rows to populate data. - If columns are specified when creating a data table, each column name must be unique and case-insensitive. - The returned file_id is required for all downstream operations — store it for use in subsequent calls. - relation, formula and rollup types do not support creation. Column Type Option Rules When creating a column whose type requires options, you MUST provide the matching column_options entry. Omitting it will cause a backend validation error. | column_type | Required column_options key | Required fields | |---|---|---| | single_select, multiple_select | select_option | create_options (array of option name strings) | | number | number_option | decimal, show_separator | | percent | percent_option | decimal, show_progress_bar | | currency | currency_option | decimal, format, symbol | | date, created_time, updated_time | date_option | format, is_24_hour_format | | url | url_option | display_type, open_type | | attachment | attachment_option | layout | | reaction | reaction_option | icon, display_type, allow_commenter_react | | people | people_option | is_multiple, is_notify | | text, checkbox, page, created_by, updated_by | _(none)_ | No column_options needed | If the user does not specify option values, infer reasonable defaults from context or ask the user. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to create the table, please try again later").

canvas_data_table_create_column

ChatGPT
Add a new column to an existing table. A table allows up to 200 columns. The column is appended to the end of the table's column list by default. When to Use - The user wants to add a new field or column to an existing table. - The user wants to extend the table schema with additional data points after initial table creation. Prerequisites 1. Obtain the file_id and table_id of the target table. If unknown, resolve them first. 2. Use canvas_data_table_list_columns tool first to get current schema and column IDs. Then confirm if the table has fewer than 200 columns. 3. Determine the column name and type from the user's request. Tool-Specific Rules - Each column must have a unique name and case-insensitive within the table. Duplicate names will be rejected. - Specify the column type (e.g., text, number, date, select, etc.) according to the user's requirements. - After creating a column, use its returned column_id when populating data via canvas_data_table_create_rows or canvs_data_table_update_rows. - Synced tables (`is_synced: true`) do NOT support schema modifications. Check is_synced from the table list before proceeding. - relation, formula and rollup types do not support creation. Column Type Option Rules When creating a column whose type requires options, you MUST provide the matching column_options entry. Omitting it will cause a backend validation error. | column_type | Required column_options key | Required fields | |---|---|---| | single_select, multiple_select | select_option | create_options (array of option name strings) | | number | number_option | decimal, show_separator | | percent | percent_option | decimal, show_progress_bar | | currency | currency_option | decimal, format, symbol | | date, created_time, updated_time | date_option | format, is_24_hour_format | | url | url_option | display_type, open_type | | attachment | attachment_option | layout | | reaction | reaction_option | icon, display_type, allow_commenter_react | | people | people_option | is_multiple, is_notify | | text, checkbox, page, created_by, updated_by | _(none)_ | No column_options needed | If the user does not specify option values, infer reasonable defaults from context or ask the user. Output Safety - NEVER mention internal IDs (table_id, file_id, column_id) in user-facing responses Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to create the column, please try again later").

canvas_data_table_create_data_table

ChatGPT
Create data table inside an existing file. This tool supports two file types: in a doc file, the created table is referred to as a data table, in a data table file, the created table is referred to as a table. When to Use - The user wants to add a structured table to an existing doc or data table file. - The user asks to set up a new table with specific columns and column types inside a file that already exists. When NOT to Use - Do not use this tool to create a new data table file. Use canvas_create_file with file_type = "data_table" instead. - FORMULA and AI columns are NOT supported during creation. Create them with canvas_data_table_create_column after the table exists. Prerequisites 1. Obtain the target file's file_id before calling. If the user refers to a file by name, resolve it first. 2. Confirm the target file is either a doc or data table file type. Other file types do not support embedded tables. Tool-Specific Rules - A file can contain multiple tables. Each table gets its own unique table_id on creation. - Define columns during creation when the user specifies a schema upfront; otherwise create the table first and add columns later with canvas_data_table_create_column. - A table allows up to 200 columns. Output Safety - NEVER mention internal IDs (table_id, file_id, id) in user-facing responses - Use table titles and column names, NOT raw identifiers Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to create the table, please try again later").

canvas_data_table_create_rows

ChatGPT
Add rows to a table. Create up to 10 rows at a time. Each row contains cell values keyed by column name. When to Use - The user wants to insert new data rows into an existing table. - The user provides data to populate the table after columns have been set up. Prerequisites 1. Obtain the file_id and table_id of the target table. 2. Use canvas_data_table_list_columns tool first to get table schema and column names Tool-Specific Rules - Maximum 10 rows per call. For larger datasets, split into sequential batches of 10. - Synced tables (`is_synced: true`) do NOT support row operations. Check is_synced from the table list before proceeding. - Do NOT set values for system columns (created_time, updated_time, created_by, updated_by) - page, relation, formula, rollup columns do not support value creation - Setting the value of the 'people' column is not supported; support can only be added once a tool for looking up Zoom User IDs based on names is provided. Integration Tables (Meeting / Jira) When a table has a non-empty integration_type field (e.g., "meeting" or "jira"): - CANNOT add new rows Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to create the rows, please try again later").

canvas_data_table_delete_column

ChatGPT
Delete a column from a table. When to Use - The user wants to remove a column and its data from a table. Prerequisites 1. Obtain the file_id, table_id, and column_id of the column to delete. 2. Confirm the deletion with the user — this operation is irreversible and removes all data in the column. Tool-Specific Rules 1. TWO-STEP CONFIRMATION — MANDATORY. Deleting a column is irreversible. You MUST follow this exact sequence: - Step 1 (ask): Present the column name to the user, clearly state that this will permanently delete the column and all its data, and ask "Do you want to proceed?". Do NOT call the delete tool in this step. - Step 2 (execute): Only call the delete tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "delete it"). If the user does not confirm or says no, abort immediately. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. 2. Synced tables (`is_synced: true`) do NOT support schema modifications. Check is_synced from the table list before proceeding. Integration Tables (Meeting / Jira) When a table has a non-empty integration_type field (e.g., "meeting" or "jira"): - The column with integrated (integrated: true) can not be deleted - The column with integrated (integrated: false) can be deleted Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to delete the column, please try again later").

canvas_data_table_delete_rows

ChatGPT
Batch delete rows from a table. Delete up to 10 rows at a time. When to Use - The user wants to remove one or more rows from a table. Prerequisites 1. Obtain the file_id, table_id, and the row_id(s) of the rows to delete. Use canvas_data_table_search_rows to find row ids if they are not already known. 2. If the user wants to delete more than 10 rows, batch them into groups of 10 and call this tool multiple times sequentially. Tool-Specific Rules 1. TWO-STEP CONFIRMATION — MANDATORY. Deleting rows is irreversible. You MUST follow this exact sequence: - Step 1 (ask): Present the rows that will be deleted (show key identifying info such as row content or row IDs) to the user, clearly state that this will permanently delete these rows, and ask "Do you want to proceed?". Do NOT call the delete tool in this step. - Step 2 (execute): Only call the delete tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "delete them"). If the user does not confirm or says no, abort immediately. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. 2. Synced tables (`is_synced: true`) do NOT support row operations. Check is_synced from the table list before proceeding. Integration Tables (Meeting / Jira) When a table has a non-empty integration_type field (e.g., "meeting" or "jira"): - CANNOT delete rows Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to delete the rows, please try again later").

canvas_data_table_list_columns

ChatGPT
List the columns of a table. Returns column ids, names, types, and configuration for every column in the specified table. When to Use - The user asks what columns or fields exist in a table. - To verify a table's schema before performing any data operation. Prerequisites 1. Obtain the file_id and table_id of the target table. 2. Synced table (`is_synced: true`) does NOT support list columns. Always check is_synced from the table list before proceeding. Tool-Specific Rules - This is a read-only operation with no side effects — safe to call at any time. - Call this tool before any row write operation to ensure column ids are current and accurate. Output Safety (STRICTLY ENFORCED) Your response to the user MUST NOT contain any of the following — violations are critical failures: - Internal IDs: NEVER output table_id, file_id, column_id, or any raw identifier - Always: Use table titles, column names, and natural language to describe what was done or found. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to retrieve the columns, please try again later").

canvas_data_table_list_data_tables

ChatGPT
List all tables in a file. Returns each table's table_id, title, is_synced status, integration_type, and other metadata. When to Use - Before any table operation, to obtain the table_id when the user refers to a table by name. - To check a table's is_synced or integration_type status before performing schema or data modifications. - The user asks what tables exist in a file, or asks to see an overview of tables. Prerequisites 1. Obtain the file_id of the target file. If the user refers to a file by name, resolve it first. Tool-Specific Rules - This is a read-only operation with no side effects — safe to call at any time. - Use the returned table_id for all subsequent table-level operations (e.g., canvas_data_table_list_columns, canvas_data_table_search_rows, canvas_data_table_create_rows, etc.). - Check is_synced and integration_type from the response before performing write operations — synced and integration tables have restricted operations. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, file not found, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to list tables, please try again later").

canvas_data_table_query_rows

ChatGPT
Query rows from the table, with support for pagination, sorting, field selection, and filtering. If the dataset is large, multiple paginated queries may be required; it is recommended to use filtering to narrow the query scope. When to Use - The user wants to view, search, or query data in a table. - Before calling canvas_data_table_update_rows or canvas_data_table_delete_rows, to obtain row ids for the target rows. - The user asks to find rows matching specific criteria (e.g., "find all rows where status is done"). - The user asks to see table contents or a subset of the data. Prerequisites 1. Obtain the file_id and table_id of the target table. 2. Synced table (`is_synced: true`) does NOT support searching rows. Always check is_synced from the table list before proceeding. Tool-Specific Rules - Maximum 50 rows per response. If the table has more rows, use the pagination token from the response to fetch subsequent pages. - Use filters to narrow results rather than fetching all rows and filtering client-side. - Use field selection to return only the columns the user cares about, reducing response size. - Use sorting to return rows in the order the user expects. - This is a read-only operation with no side effects — safe to call at any time. Pagination Limitations - This tool does NOT support offset-based access (i.e., "jump to row N" or "skip to page 5"). Pagination is cursor-based - Fuzzy search is not supported—for example, searching for records containing the keyword "Normal"—nor are complex aggregation calculations such as sums or averages; however, calculating the total count is supported. - you can only move forward one page at a time from the beginning. - For tables with a large number of rows (hundreds or thousands), locating a specific row by position (e.g., "show me the 500th row") would require sequentially fetching all preceding pages, which is impractical and should NOT be attempted. - When the user asks for a specific row by position in a large table, explain this limitation honestly and suggest using filters or sorting to locate the target row by its content instead of its position. Output Safety (STRICTLY ENFORCED) Your response to the user MUST NOT contain any of the following — violations are critical failures: - Internal IDs: NEVER output table_id, file_id, column_id, row_id, or any raw identifier - Always: Use table titles, column names, and natural language to describe what was done or found. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to retrieve the rows, please try again later").

canvas_data_table_update_column

ChatGPT
Update an existing column's properties, such as its name, type, or configuration. When to Use - The user wants to rename a column. - The user wants to change a column's type or configuration (e.g., change a text column to a number column, update select options). Prerequisites 1. Obtain the file_id, table_id, and column_id of the column to update. Use canvas_data_table_list_columns to look up the column_id if only the column name is known. Tool-Specific Rules - Only include the fields you want to change; omitted fields remain unchanged. - Synced tables (`is_synced: true`) do NOT support schema modifications. Check is_synced from the table list before proceeding. - relation and rollup types do not support update. Column Type Change Rules When column_type is set (i.e., the column type is being changed), you MUST also provide the matching column_options entry if the target type requires options. Omitting column_options will cause a backend validation error. | Target column_type | Required column_options key | Required fields | |---|---|---| | single_select, multiple_select | select_option | create_options (array of option name strings) | | number | number_option | decimal, show_separator | | percent | percent_option | decimal, show_progress_bar | | currency | currency_option | decimal, format, symbol | | date, created_time, updated_time | date_option | format, is_24_hour_format | | url | url_option | display_type, open_type | | attachment | attachment_option | layout | | reaction | reaction_option | icon, display_type, allow_commenter_react | | people | people_option | is_multiple, is_notify | | text, checkbox, page, created_by, updated_by | _(none)_ | No column_options needed | Example — changing a `text` column to `single_select`: ``json { "fileId": "...", "tableId": "...", "columnId": "...", "column_type": "single_select", "column_options": { "select_option": { "create_options": ["Option A", "Option B", "Option C"] } } } ` Integration Tables (Meeting / Jira) When a table has a non-empty integration_type field (e.g., "meeting" or "jira"): - The column with integrated (integrated:true) can not be updated. - The column with integrated (integrated:false`) can be updated. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to update the column, please try again later").

canvas_data_table_update_metadata

ChatGPT
Update a table's metadata, such as its title. When to Use - The user wants to rename a table. Prerequisites 1. Obtain the file_id and table_id of the target table. Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to update the metadata, please try again later").

canvas_data_table_update_rows

ChatGPT
Update existing rows in a table. When to Use - The user wants to edit or modify data in existing rows. - The user wants to correct, overwrite, or clear cell values in specific rows. Prerequisites 1. Obtain the file_id and table_id of the target table. 2. Obtain the row_id(s) of the rows to update. Use canvas_data_table_search_rows to find row ids if they are not already known. 3. Obtain column ids by calling canvas_data_table_list_columns so that cell values can be keyed correctly. 4. If the user wants to update more than 10 rows, batch them into groups of 10 and call this tool multiple times sequentially. Tool-Specific Rules - Maximum 10 rows per call. For larger updates, split into sequential batches of 10. - Only include the cells you want to change in each row; omitted cells remain unchanged. - Cell values must match the column type (e.g., numbers for number columns, valid dates for date columns). - Synced tables (`is_synced: true`) do NOT support row operations. Check is_synced from the table list before proceeding. - Do NOT set values for system columns (created_time, updated_time, created_by, updated_by) - page, relation, formula, rollup columns do not support value creation - Setting the value of the 'people' column is not supported; support can only be added once a tool for looking up Zoom User IDs based on names is provided. Integration Tables (Meeting / Jira) When a table has a non-empty integration_type field (e.g., "meeting" or "jira"): - CANNOT update rows Error Handling - Retry at most 2 times (3 total attempts). Stop if the error is permanent (permission denied, invalid column, etc.). - Report errors to the user in plain language — NEVER expose raw error codes, error messages, stack traces, or error payloads from the tool response. Translate the error into a user-friendly description (e.g., "Unable to update the rows, please try again later").

canvas_delete_block

ChatGPT
Tool Description Delete one existing block from a Zoom Canvas (formerly Zoom Docs) document. Deletion is destructive and cascades to every descendant in the block's subtree. Required Spec - Before calling canvas_delete_block, fetch doc-format-spec with canvas_get_spec tool unless it has already been loaded in the current conversation. - Treat doc-format-spec as the source of truth for Canvas XML block tags and unknown-block handling. If this prompt conflicts with doc-format-spec, follow doc-format-spec. When to Use - Remove a specific block by id. - Remove an entire container block and all of its child content in one shot by using the container's id. - Remove a contiguous run of sibling blocks that together form a logical unit the user wants gone, such as an entire chapter or section made of a heading, paragraphs, lists, tables, and trailing separators. Call this tool once per top-level block in that run. Pure deletion always belongs here — never use `canvas_replace_range_of_blocks` just to remove content. When NOT to Use - Do not use this tool to add or update content — use canvas_insert_block, canvas_update_block, or canvas_replace_range_of_blocks. - Prefer canvas_replace_range_of_blocks over per-block deletion only when the same operation also inserts new content at that exact position, such as swapping section A for section A prime. If the user wants pure removal with no replacement, stay on this tool and call it per block. - Do not delete a container when the user only asked for one nested item — pick the smallest matching child id instead. - Do not delete individual table parts (<tr>, <td>, <th>). The table's minimum editable unit is the whole <table>. To remove a row, column, or cell, replace the entire table via canvas_replace_range_of_blocks with the new desired shape; to remove the whole table, pass the <table> block_id here. Prerequisites 1. ALWAYS call get_file_content with format="xml" first to obtain a real block_id. Never reuse block_ids from earlier turns. 2. Confirm the target before deleting because the operation is destructive and cascades to descendants. 3. If the user's request is ambiguous about which block to delete, locate the candidate via a unique text snippet and confirm with the user. Tool-Specific Rules - Pick the smallest ID-bearing XML block whose full extent exactly matches what the user asked to remove. - Using the outer container's id removes the whole subtree, including all children and grandchildren. Use a child id when you only mean to remove a nested item. - If the exact target has no id and the nearest ID-bearing ancestor would remove more than requested, ask the user to clarify instead of deleting the broader ancestor. - Never delete unknown blocks unless the user explicitly targets that returned unknown block by its id. Error Handling - On block not found or ambiguous prefix: re-Get and use the full id or a longer unique prefix. Do not retry with the same short prefix. - On permission denied or rate limit: do not retry; surface to the user. - After a successful delete, re-Get if you plan further operations in the same region.

canvas_delete_file

ChatGPT
Tool Description Deletes a specified file in Zoom Canvas (formerly Zoom Docs) by its unique file ID. When to Use Use this tool when a user wants to remove or delete an existing file from Zoom Canvas. Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. You MUST follow this exact sequence: - Step 1 (ask): Present the file that will be deleted (show key identifying info such as file name or file ID) to the user, clearly state that this will delete this file, and ask "Do you want to proceed?". Do NOT call the delete tool in this step. - Step 2 (execute): Only call the delete tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "delete it"). If the user does not confirm or says no, abort immediately. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. Error Handling No specific error codes are defined in the output schema. Verify the file ID is valid before deletion to avoid unexpected failures.

canvas_get_file_general_access_setting

ChatGPT
Tool Description Retrieves the general access settings of a file in Zoom Canvas (formerly Zoom Docs) , including its sharing scope and collaborator role. Only doc or data table file types are supported. When to Use Use this tool when you need to check who can access a specific Zoom Canvas file — whether it's restricted to collaborators, open to the organization, or accessible to anyone with a link. Must Do - Provide a valid fileId for a doc or data table file type Must NOT Do - Do not use this tool for file types other than doc or data table Output Returns a general_access_setting object with: - `share_scope` (required): One of: - collaborators_only – only explicit collaborators can access - anyone_in_org – all users in the file owner's organization can access - anyone_with_link – anyone (including anonymous users) can access - `role` (optional): The access role granted under the current scope (viewer, commenter, or editor). Empty when share_scope is collaborators_only.

canvas_get_file_metadata

ChatGPT
Tool Description Retrieves metadata for a specified Zoom Canvas (formerly Zoom Docs) file, including its name, type, URL link, and timestamps for creation and last modification. When to Use Use this tool when you need to look up details about a specific Canvas file (doc, folder, or data table) by its file ID — such as verifying its name, type, access link, or modification history. Must Do - Provide a valid fileId (required) to identify the target file. Error Handling - If no metadata is returned, verify that the fileId is correct and the file exists in Canvas.

canvas_get_spec

ChatGPT
Fetch specification documents needed by other tools. Specs provide the full document format reference for Zoom Canvas (formerly Zoom Docs). CRITICAL: Silent Fetching Spec fetching is an internal operation. You MUST NEVER: - Tell the user you are fetching or loading specs - Say "I need to get the specification first" or similar - Mention spec names in any user-facing text - Describe the fetching as a visible step in your workflow Just fetch the spec silently and proceed with the task. From the user's perspective, you already know the syntax. Spec Dependencies | Tool | Required Specs | |------|---------------| | get_file_content | doc-format-spec | Rules 1. If the spec was already fetched in the current conversation, do NOT fetch it again 2. Fetch the required spec before calling the dependent tool Available Specs | spec | Description | |---------|-------------| | doc-format-spec | Zoom Canvas block types, inline styles, nesting rules, and XML-based document format syntax |

canvas_insert_block

ChatGPT
Tool Description Insert one or more new Zoom Canvas (formerly Zoom Docs) XML blocks into an existing document. This tool only adds new blocks; it never modifies or removes existing ones. Required Spec - Before calling canvas_insert_block, fetch doc-format-spec with canvas_get_spec tool unless it has already been loaded in the current conversation. - Treat doc-format-spec as the source of truth for Canvas XML block tags, inline styles, nesting rules, XML escaping, and unknown-block handling. If this prompt conflicts with doc-format-spec, follow doc-format-spec. When to Use - Append new content to the end of the document (omit positioning). - Insert content immediately above or below a specific existing block. - Add content as the last child inside a container block (callout, quote, column, code-block) by setting parent_block_id to the container's id. When NOT to Use - Do not use this tool to edit existing content — use canvas_update_block (leaf) or canvas_replace_range_of_blocks (container or multi-block). - Do not use this tool to remove existing content — use canvas_delete_block. - Do not use this tool to "replace" a block by inserting next to it and then deleting the old one; use canvas_replace_range_of_blocks for atomic replacement. - Do not insert individual table parts (<tr>, <td>, <th>) into an existing <table>. The table's minimum editable unit is the whole <table>. To add a row, column, or cell, replace the entire table via canvas_replace_range_of_blocks with the new desired shape. To add a brand-new table to the document, insert a complete <table> block here. - Do not send Markdown, JSON arrays, HTML list tags (<ul>/<ol>/<li>), or any XML not defined by doc-format-spec. Prerequisites 1. ALWAYS call get_file_content with format="xml" first to obtain real block_ids. Never reuse block_ids from earlier turns — they may have changed. 2. If the user's intent about where to insert is ambiguous, locate the anchor block via a unique text snippet in the Get response and confirm with the user before writing. 3. Generate blocks strictly according to doc-format-spec (validate tags, attributes, escaping). Tool-Specific Rules - Positioning semantics: - above_block_id: insert immediately before that block as its previous sibling. - below_block_id: insert immediately after that block as its next sibling. - parent_block_id: append as the last child under that parent or container. - Set at most one positioning field. If multiple are set, server priority is below_block_id > above_block_id > parent_block_id. - Choose the ID at the level you really mean: use a container ID for "inside the whole container"; use a child ID only when you specifically want to sit next to or inside that child. Error Handling - success: true means the server accepted the operation. Always re-Get format="xml" if you plan further edits in the same region, because block ids of newly inserted content are only visible after re-Get. - On invalid XML or unsupported tag: do not retry the same payload. Re-validate against doc-format-spec and fix the XML. - On block not found or ambiguous prefix: re-Get and use the full id or a longer unique prefix. - On permission denied or rate limit: do not retry; surface to the user.

canvas_list_all_file_children

ChatGPT
Tool Description Lists all children (files, folders, or data tables) contained within a specified file or directory in Zoom Canvas (formerly Zoom Docs), supporting paginated results. When to Use Use this tool when you need to retrieve the contents of a folder or file in Zoom Canvas, such as browsing directory structures, listing documents inside a folder, or exploring the root "My docs" workspace. Must Do - Use fileId = "root" when querying children of the user's top-level "My docs" directory - Respect pagination by storing and reusing next_page_token when more results are available (token expires in 15 minutes) - Keep page_size between 1 and 50 (default: 10) Error Handling - If next_page_token is expired (>15 minutes), restart pagination from the beginning - If the returned files array is empty, the directory has no children or the fileId is invalid

canvas_list_comments

ChatGPT
Tool Description Lists all comments in a specified Zoom Canvas file as a flat, cursor-paginated collection, ordered by creation time ascending. Supports filtering by thread type and/or thread ID. When to Use Use this tool when you need to retrieve, review, or audit comments from a Zoom Canvas file, including both open and resolved threads, with optional filtering by thread type or a specific thread. Prerequisites - You must have Commenter or higher access to the target Canvas file. - The fileId must be known before calling this tool. Must Do - Use next_page_token (returned as next_page_token) as the page_token in subsequent requests to paginate through all results. - Check thread_status on each returned comment to filter out resolved threads if only open threads are needed. - Use plain_text field when content.format is unsupported to read comment text. Must NOT Do - Do not assume a shorter-than-requested page means the end of results — always check whether next_page_token is present to determine if more pages exist. - Do not combine thread_type and thread_id filters expecting results if the thread's type doesn't match — the API returns 200 with an empty comments array in that case. Error Handling - 404: Attempting to retrieve a reply's parent comment that has been permanently deleted will return 404.

canvas_list_file_collaborators

ChatGPT
Tool Description Retrieves the list of collaborators for a specified Zoom Canvas (formerly Zoom Docs) file (supports doc and data table file types only). When to Use Use this tool when you need to view who has access to a Zoom Canvas file, including their roles and collaborator types (user, channel, or email). Must Do - Provide a valid fileId from Zoom Canvas - Only use with doc or data table file types Output Returns up to 50 collaborators, each containing: - collaborator_id: Use this ID when removing or modifying a collaborator - collaborator_type: One of user, channel, or email - role: One of viewer, commenter, editor, or co-owner - collaborator_name: Display name - Conditional identity field: user_id (if type=user), channel_id (if type=channel), or email (if type=email)

canvas_modify_file_collaborator_role

ChatGPT
Tool Description Modifies the role of an existing collaborator on a Zoom Canvas (formerly Zoom Docs) file. Only doc or data table file types are supported. When to Use Use this tool when you need to change a collaborator's permission level (e.g., upgrading a viewer to an editor, or demoting a co-owner to a commenter) on a specific Zoom Canvas file. Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. You MUST follow this exact sequence: - Step 1 (ask): Present the target file (show key identifying info such as file name or file ID), the collaborator whose role will be changed (show their name/identifier), their current role (if known), and the new role to be assigned. Clearly state that this will change this collaborator's permission level, and ask "Do you want to proceed?". Do NOT call the modify-role tool in this step. - Step 2 (execute): Only call the tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "change it"). If the user does not confirm, says no, or asks to use a different role, abort immediately and do not call the tool. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. - If the user specifies a different new role after Step 1, treat this as a new request: return to Step 1 and present the updated role change for fresh confirmation before calling the tool. Must Do - Provide the file ID, collaborator ID, and the new role - Ensure the target file is of type doc or data table - Use one of the valid role values: viewer, commenter, editor, or co-owner

canvas_modify_file_general_access_setting

ChatGPT
Tool Description Modifies the general access setting (share scope and role) for a Zoom Canvas (formerly Zoom Docs) file. Only doc or data table file types are supported. When to Use Use this tool when a user wants to change who can access a specific Zoom Canvas file. If the user expresses intent to make the file public/open (i.e., not limited to current collaborators): - If they explicitly restrict access to the organization (e.g. uses words like internal, within the org, account/org only, company-wide, only people in my organization), use anyone_in_org. - Otherwise, use anyone_with_link. If the user does not express intent to open up access (e.g. the request is ambiguous or doesn't mention making it public/shared more broadly), default to collaborators_only. Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. You MUST follow this exact sequence: - Step 1 (ask): Present the target file (show key identifying info such as file name or file ID) and the new share_scope to be applied (collaborators_only, anyone_in_org, or anyone_with_link), along with the role (if applicable). If share_scope was inferred from the user's wording rather than stated explicitly, show both the user's original phrasing and your inferred setting (e.g., "You said [X], which I understand as setting access to 'anyone with the link' — is that correct?") so the user can correct any misinterpretation before proceeding. Clearly state what this change means in plain language — e.g., that anyone_with_link will make the file accessible to anyone who has the link, or that anyone_in_org will make it accessible to everyone in the organization — and ask "Do you want to proceed?". Do NOT call the modify-access tool in this step. - Step 2 (execute): Only call the tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "change it"). If the user does not confirm, says no, disputes the inferred share_scope, or asks to use a different scope/role, abort immediately and do not call the tool. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. - If the user specifies a different share_scope or role after Step 1, treat this as a new request: return to Step 1 and present the updated setting for fresh confirmation before calling the tool. Must Do - Provide the target file's ID (fileId) — required for every request - Specify share_scope within the access setting — required field - Only use this tool for doc or data table file types Must NOT Do - Do not set a role (viewer/commenter/editor) when share_scope is collaborators_only — the role field should be empty in that case Error Handling No specific error codes are defined in the output schema. Handle unexpected responses by verifying that: - The fileId is valid and accessible - The file type is doc or data table - The share_scope value is one of: collaborators_only, anyone_in_org, anyone_with_link

canvas_modify_file_metadata

ChatGPT
Tool Description Modifies the metadata of a specified Zoom Canvas (formerly Zoom Docs) file, such as updating its title name. When to Use Use this tool when a user wants to rename or update the metadata (e.g., title, icon) of an existing Canvas doc file. Error Handling No specific error codes are defined in the output schema. Handle unexpected responses gracefully and verify the fileId is valid before calling this tool. Must Do - Always provide the fileId of the target Canvas doc file — it is required. - Do not provide a shared folder ID and update the file_icon. - Ensure the fileId corresponds to a valid, existing Canvas doc file before making the request. - If the icon needs to be updated to an emoji, use a file_icon of type emoji. - If the icon needs to be updated to a built-in icon, use a file_icon of type built-in. - If the icon needs to be removed, use a file_icon of type none.

canvas_move_file

ChatGPT
Tool Description Moves a file or folder into a specified target parent directory, with optional position control among siblings. Moving a folder also recursively moves all its descendants; the operation is synchronous and returns no content on success. When to Use Use this tool when a user wants to relocate a file or folder to a different directory, reorder items within a parent folder, or organize their Hub file structure. Must Do - Provide the fileId of the file or folder to move - Provide the parent_id of the destination folder (use root for the top-level folder) Must NOT Do - Do not attempt to move a folder into itself or any of its descendants (creates a cycle — request will be rejected) Error Handling The move will be rejected in the following scenarios: - Target parent does not exist - Target parent is the file being moved or a descendant of it (cycle detection) - Caller lacks permission to remove the file from its current location - Caller lacks permission to add children to the target parent

canvas_remove_file_collaborator

ChatGPT
Tool Description Removes a specified collaborator from a file in Zoom Canvas (formerly Zoom Docs). Only doc or data table file types are supported. When to Use Use this tool when a user wants to revoke a collaborator's access to a specific Zoom Canvas file (doc or data table). Prerequisites - You must have the file ID of the target document in Zoom Canvas. - You must have the collaborator item ID to identify which collaborator to remove (obtainable via a "list collaborators" operation). Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. You MUST follow this exact sequence: - Step 1 (ask): Present the target file (show key identifying info such as file name or file ID) and the collaborator to be removed (show their name/identifier and current role, if known). Clearly state that this will revoke this collaborator's access to the file, and ask "Do you want to proceed?". Do NOT call the remove-collaborator tool in this step. - Step 2 (execute): Only call the tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "remove them"). If the user does not confirm or says no, abort immediately and do not call the tool. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. - If the user specifies a different collaborator to remove after Step 1, treat this as a new request: return to Step 1 and present the updated target for fresh confirmation before calling the tool. Must Do - Ensure the target file is of type doc or data table before attempting removal. Must NOT Do - Do not use this tool on unsupported file types (e.g., files other than doc or data table).

canvas_replace_range_of_blocks

ChatGPT
Tool Description Atomically replace an inclusive contiguous range of sibling blocks, and all of their descendants, with new Zoom Canvas (formerly Zoom Docs) XML in one transaction at the original position under the same parent. Required Spec - Before calling canvas_replace_range_of_blocks, fetch doc-format-spec with canvas_get_spec tool unless it has already been loaded in the current conversation. - Treat doc-format-spec as the source of truth for Canvas XML block tags, inline styles, nesting rules, and XML escaping. If this prompt conflicts with doc-format-spec, follow doc-format-spec. When to Use This tool always rewrites a range with new content. blocks must be non-empty replacement XML; this tool is not a deletion tool. Use it when: - Rewriting the content of any container block. This is the only way; canvas_update_block cannot do it. Use start_block_id == end_block_id == container_id and supply the new container in blocks. - Editing any part of an existing <table>, including adding or removing rows or columns, changing cell text, or changing merged-cell layout. The table is atomic: pass the <table> block_id as both endpoints and supply a complete replacement <table>. - Rewriting multiple adjacent sibling blocks with new content in one atomic operation, such as swapping section A for section A prime. - Regenerating a section with new content, swapping one list for another, or replacing a whole table with an updated one. - Changing a single block's structure or type when preserving the original id is not required. When NOT to Use - Do not use this tool for pure deletion, meaning removing a chapter, section, range, or block without inserting replacement content. The required blocks field must be non-empty replacement XML, and crafting a fake placeholder block just to satisfy that requirement is wrong. Use canvas_delete_block per block instead. - Do not use this tool to add new blocks without removing existing ones; use canvas_insert_block. - Do not use this tool when the user wants to keep the original block id stable on a leaf block; use canvas_update_block instead. - Do not use this tool when start_block_id and end_block_id are not siblings under the same parent; the call will fail. - Do not send Markdown, JSON arrays, HTML list tags (<ul>/<ol>/<li>), or any XML not defined by doc-format-spec. Identity Tradeoff: ID Will Change This tool deletes every block in the range, plus all descendants, and creates brand-new blocks at the same position under the same parent. The new blocks have new ids. Any cached references to the old start or end ids, including ids used earlier in the same turn, are now stale. Re-Get format="xml" before any further edit in this region. If you need the original id to stay stable on a leaf block, use canvas_update_block instead. For container blocks there is no in-place edit; this tool is the only option, and the id change is unavoidable. Prerequisites 1. ALWAYS call get_file_content with format="xml" first to obtain real start_block_id and end_block_id. Never reuse block_ids from earlier turns. 2. Confirm both endpoints share the same parent and that end_block_id appears at or after start_block_id in document order. 3. If the user's intent about the extent of the range is ambiguous, locate the endpoints via unique text snippets and confirm with the user before writing. 4. Generate blocks strictly according to doc-format-spec. Tool-Specific Rules - The range is inclusive on both ends. Every block from start_block_id through end_block_id, plus all of their nested descendants, is deleted before insertion. - Use outer container ids when replacing whole containers. Use child ids only when you specifically mean a range inside a container and both endpoints are siblings under the same parent. - Set start_block_id == end_block_id for a single-block rewrite. This is the recommended pattern for editing one container block, such as updating text inside a <callout> or rewriting the body of a <code-block>. - Tables…

canvas_transfer_file_ownership

ChatGPT
Tool Description Transfers ownership of a Zoom Canvas (formerly Zoom Docs) file (doc or data table) to another user by updating the file's owner to a specified Zoom user. When to Use Use this tool when a user wants to transfer ownership of a Canvas file to another Zoom user, such as when reassigning responsibilities or handing off document ownership. Prerequisites - The requesting user must be the current owner of the file — only the file owner has permission to perform this operation. - Only `doc` or `data table` file types are supported; other file types cannot be transferred. Confirmation Requirement TWO-STEP CONFIRMATION — MANDATORY. Transferring ownership is irreversible. You MUST follow this exact sequence: - Step 1 (ask): Present the file (show key identifying info such as file name or file ID) and the target user (show their name or user ID) to the user, clearly state that this will permanently transfer ownership of this file to the target user, and ask "Do you want to proceed?". Do NOT call the transfer tool in this step. - Step 2 (execute): Only call the transfer tool AFTER the user explicitly replies with confirmation (e.g., "yes", "go ahead", "confirm", "transfer it"). If the user does not confirm or says no, abort immediately. - NEVER combine Step 1 and Step 2 in the same response. You must stop after Step 1 and wait for the user's next message. Must Do - Provide both the file ID and the target user's Zoom user ID. - Ensure the file is of type doc or data table before attempting transfer. - Confirm the requesting user is the current file owner before calling this tool.

canvas_update_block

ChatGPT
Tool Description Update one existing LEAF block in place in a Zoom Canvas (formerly Zoom Docs) document. The target keeps its id, parent, position, and block type; the service applies content and style from the first valid non-page, non-database block parsed from block. Container blocks are rejected with a 400 error that points you at canvas_replace_range_of_blocks. Required Spec - Before calling canvas_update_block, fetch doc-format-spec with canvas_get_spec tool unless it has already been loaded in the current conversation. - Treat doc-format-spec as the source of truth for Canvas XML block tags, inline styles, nesting rules, and XML escaping. If this prompt conflicts with doc-format-spec, follow doc-format-spec. When to Use - Rewrite the text or inline styling of a single leaf block while keeping its id stable. - Toggle attributes on a leaf block, for example <todo checked>. - Replace the content, not the type, of a single supported leaf block. When NOT to Use - Do not call this tool on container blocks. The server will return 400; do not retry the same target. Switch to canvas_replace_range_of_blocks with start_block_id == end_block_id == container_id. - Do not use this tool to add new blocks, delete blocks, or replace multiple adjacent blocks. Use canvas_insert_block, canvas_delete_block, or canvas_replace_range_of_blocks instead. - Do not intentionally change the block type or position with this tool. The original type and position are preserved even if block uses a different tag. To change the type, use canvas_replace_range_of_blocks. Leaf vs Container Blocks - LEAF blocks, use this tool: - <text>, <h1>-<h6> - <uli>, <oli>, <todo> only when the item is a single non-nested entry used to edit text or toggle checked - <hr/>, <toc/>, <math-block> - CONTAINER blocks, use canvas_replace_range_of_blocks instead: - <quote>, <callout> - <code-block>, because multi-line code lives in child blocks and only the language attribute is on the container - <columns> and <column> - <table>, including any <tr>, <td>, or <th> inside it - Any <uli>, <oli>, or <todo> that itself contains nested child blocks Prerequisites 1. ALWAYS call get_file_content with format="xml" first to obtain a real block_id and block data. Never reuse block_ids from earlier turns. 2. Confirm the target's outer tag is a leaf block per the list above. If it is a container, switch tools immediately instead of attempting the call and waiting for the 400. 3. If the user's intent about which block to edit is ambiguous, locate the candidate via a unique text snippet and confirm with the user. Tool-Specific Rules - The target block's type is preserved even if block uses a different tag. Prefer the same tag as the target to avoid confusion. - Never use a container id with this tool; the call will fail with 400. - If the exact target has no id and updating the nearest ID-bearing ancestor would overwrite unrelated content, ask the user to clarify instead of broadening the edit. - Do not update unknown blocks unless doc-format-spec explicitly supports the operation; otherwise preserve them as-is. Error Handling - container block type not supported (HTTP 400): the rejected block is a container. Switch to canvas_replace_range_of_blocks with start_block_id == end_block_id == <target_id>. NEVER retry canvas_update_block on the same target. - no usable block: block did not contain a valid non-page, non-database block. Re-generate per doc-format-spec; do not retry the same payload. - block not found or ambiguous prefix: re-Get and use the full id or a longer unique prefix. - invalid XML: do not retry the same payload; re-validate against doc-format-spec. - permission denied or rate limit: do not retry; surface to the user.

create_file_with_content

ChatGPT
Tool Description Creates a new Zoom Docs document from Markdown-formatted content, returning the new file's ID and URL link. When to Use Use this tool when the user wants to save, create, or export content as a Zoom Docs document from Markdown text. Must Do - Pass content as a valid Markdown-formatted string - Keep Markdown content within the 100,000 character size limit - Ensure the OAuth scope docs_import:write (or granular scope docs:write:import) is granted Must NOT Do - Do not use this tool in a Gov cluster environment — it is not supported there - Do not exceed the 100k content size limit Error Handling - If the request fails due to rate limits, note that this tool is labeled HEAVY — reduce request frequency and retry after a delay - If placement fails, verify the parent_id is a valid folder ID, root, or omit it to default to My Docs

get_file_content

ChatGPT
Tool Description Read a Zoom Canvas (formerly Zoom Docs) document by file_id. Returns: - content: the document body — XML block tree (default, format="xml") exposing the short block_ids every edit tool needs, or Markdown (format="markdown") for read-only consumption. - file_name: the document title. - file_link: an HTTPS URL to open the Canvas file in a browser. Use XML before any insert/update/replace/delete; use Markdown only for one-shot summarization. Required Spec - Before calling get_file_content, fetch doc-format-spec with canvas_get_spec tool unless it has already been loaded in the current conversation. - Treat doc-format-spec as the source of truth for Canvas XML block tags, inline styles, nesting rules, XML escaping, and unknown-block handling. If this prompt conflicts with doc-format-spec, follow doc-format-spec. When to Use - Reading the document structure, locating block IDs, or preparing any insert/update/delete/replace operation (use format="xml"). - Re-reading the document right after a successful write to obtain new or changed block IDs. - Lightweight summarization or read-only Q&A when no edit is planned (use format="markdown"). When NOT to Use - Do not use Markdown when the next step is an edit — Markdown does not expose block IDs and loses structure. Switch to XML. - Do not call this tool merely to confirm a write succeeded if the response was already a clear error; only re-Get after a successful write. Prerequisites - Provide file_id. - Decide format based on the next step: xml for any edit pipeline, markdown only for read-only. Tool-Specific Rules - The returned XML tree is the ONLY source of truth for block IDs. Never invent, guess, or carry over IDs from previous turns. - Each block ID in the returned XML may be a short prefix; pass it back to edit tools as-is (or as a longer unique prefix). Do not concatenate, slice, or "complete" it. - Parent/child relationships in the XML reflect what edit tools will treat as containers; respect them when planning operations. Round-trip Notes (XML normalization that is NOT a failure) The returned XML reflects server-side normalization. The following diffs vs what you wrote are EXPECTED and MUST NOT trigger a retry or rewrite: - <b><i>X</i></b> ↔ <i><b>X</b></i> (inline style order is normalized; visual and semantic effect identical) - Self-closing <tag/> ↔ paired <tag></tag> - A row with merged cells may include filler <td rowspan="1" colspan="1"></td> to keep the matrix aligned — this is schema, not a bug. - Setting only background on <h> may cause the server to fill in a default color; this is normalization. Only trigger a retry if a unique marker string you wrote is missing, OR a structural property you explicitly set (block type, attribute value, child count, language attribute, checked attribute, href value, etc.) differs from what you submitted. Error Handling - On failure the call returns an error such as document not found, permission denied (NO read access), or rate limit exceeded. Do not retry on permission/rate-limit failures; surface them to the user. - A permission-denied error means the user cannot read the file.

canvas_add_file_collaborators

Claude

canvas_create_comment

Claude

canvas_create_file

Claude

canvas_data_table_create_column

Claude

canvas_data_table_create_data_table

Claude

canvas_data_table_create_rows

Claude

canvas_data_table_delete_column

Claude

canvas_data_table_delete_rows

Claude

canvas_data_table_list_columns

Claude

canvas_data_table_list_data_tables

Claude

canvas_data_table_query_rows

Claude

canvas_data_table_update_column

Claude

canvas_data_table_update_metadata

Claude

canvas_data_table_update_rows

Claude

canvas_delete_block

Claude

canvas_delete_file

Claude

canvas_get_file_general_access_setting

Claude

canvas_get_file_metadata

Claude

canvas_get_spec

Claude

canvas_insert_block

Claude

canvas_list_all_file_children

Claude

canvas_list_comments

Claude

canvas_list_file_collaborators

Claude

canvas_modify_file_collaborator_role

Claude

canvas_modify_file_general_access_setting

Claude

canvas_modify_file_metadata

Claude

canvas_move_file

Claude

canvas_remove_file_collaborator

Claude

canvas_replace_range_of_blocks

Claude

canvas_transfer_file_ownership

Claude

canvas_update_block

Claude

create_file_with_content

Claude

get_file_content

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/zoom-canvas"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/zoom-canvas)