Monday MCP

Get a monday.com form by its form token. Form tokens can be extracted from the form's url. Given a form url, such as https://forms.monday.com/forms/abc123def456ghi789?r=use1, the formToken is the alphanumeric string that appears right after /forms/ and before the ?. In the example, the formToken is abc123def456ghi789.

Ask a read-only question about an existing Vibe app. Blocks for up to 45s (configurable via timeout_ms) awaiting the assistant reply. Status: COMPLETED with the reply, TIMEOUT if the workflow did not finish in time (call vibe_get later to retrieve it), or FAILED if the workflow errored or was cancelled. Optional model to pick the LLM for the answer.

Fetch a Vibe app by id. App metadata is always returned, including editor_link : the URL of the Vibe builder/chat page for this app (https://{accountSlug}.monday.com/vibe/app/{appId}); usable as soon as the app row exists. Pass `include` to add expensive slices: status (refreshes status + adds is_busy, default true), messages (with optional from_date), code_versions.

Get information about monday.com documents. Supports two modes: MODE: "content" (default) — Fetch documents with their full markdown content. - Requires: type ("ids" | "object_ids" | "workspace_ids") and ids array - Supports pagination via page/limit. Check has_more_pages in response. - If type "ids" returns no results, automatically retries with object_ids. - Set include_blocks: true to include block IDs, types, and positions in the response — required before calling update_doc. - Blocks default to 25 per page. Use blocks_limit and blocks_page to paginate through long documents. - Set include_comments: true to fetch all comments and replies on the document. Each comment is enriched with anchor info (block_id, selection_from, selection_length) indicating which block and text range it's attached to. Use comments_limit to control how many comments per item (default 50). MODE: "version_history" — Fetch the edit history of a single document. - Requires: ids with the document's object_id (use the object_id field from content mode results, NOT the id field). - The object_id is the numeric ID visible in the document URL. - Returns restoring points sorted newest-first. Use version_history_limit to cap results (e.g., "last 3 changes" → version_history_limit: 3). - Use since/until to filter by time range. If omitted, returns full history. - Set include_diff: true to see what content changed between versions (fetches up to 10 diffs, may be slower). - Examples: - { mode: "version_history", ids: ["5001466606"], version_history_limit: 3 } - { mode: "version_history", ids: ["5001466606"], since: "2026-03-11T00:00:00Z", include_diff: true }

List Vibe apps owned by the authenticated user. Supports pagination, search, status, and is_published filters.

Create a new monday.com doc either inside a workspace or attached to an item (via a doc column). After creation, the provided markdown will be appended to the document. LOCATION TYPES: - workspace: Creates a document in a workspace (requires workspace_id, optional doc_kind, optional folder_id, optional docOwnerIds) - item: Creates a document attached to an item (requires item_id, optional column_id, optional docOwnerIds) USAGE EXAMPLES: - Workspace doc: { location: "workspace", workspace_id: 123, doc_name: "My Doc", doc_kind: "private" , markdown: "..." } - Workspace doc in folder: { location: "workspace", workspace_id: 123, doc_name: "My Doc", folder_id: 17264196 , markdown: "..." } - Item doc: { location: "item", item_id: 456, doc_name: "My Doc", column_id: "doc_col_1" , markdown: "..." } - Workspace doc with agent owner: { location: "workspace", workspace_id: 123, doc_name: "My Doc", markdown: "...", docOwnerIds: [""] }

Retrieve a saved action by ID. Example: id: "550e8400-e29b-41d4-a716-446655440000"

Get assets (files) by their IDs. Returns file metadata including name, extension, size, public URL (valid for 1 hour), thumbnail URL, upload date, and who uploaded it.

Execute a saved action by ID. Optionally pass variables (injected as environment variables, access via os.environ). Example: id: "abc-123", vars: {"board_id": 12345, "limit": 5}

[UI COMPONENT] Renders an interactive chart/graph visualization that the user can see and interact with. IMPORTANT: This is a UI DISPLAY tool - use it to RENDER visual components for the user to see and interact with. Do NOT use data-fetching tools when the user explicitly asks to "show", "display", "visualize", or "see" something visually. Use when user asks for: pie chart, bar chart, line graph, data visualization, or any graphical representation of numbers/statistics.

[UI COMPONENT] Renders an interactive table visualization that the user can see and interact with. IMPORTANT: This is a UI DISPLAY tool - use it to RENDER visual components for the user to see and interact with. Do NOT use data-fetching tools when the user explicitly asks to "show", "display", "visualize", or "see" something visually. Use when user asks to: display a board as table, show items in table format, view data in tabular layout, or see a Monday.com board visually. When asked to update an item, use the currently selected item ID (get it from the widget state, using tools like "get_widget_state") for deciding which item to update. If no item is selected, ask the user which item should be updated. After adding an update to an item, you MUST display the table AGAIN, even if the user did not ask you to. [IMPORTANT][FILTERING PRECONDITION]: IF using filters, you MUST call get_board_info(boardId) FIRST and use the returned boardContextToken.

Update an existing monday.com document. Provide doc_id (preferred) or object_id, plus an ordered operations array (executed sequentially, stops on first failure). OPERATIONS: - set_name: Rename the document. - add_markdown_content: Append markdown as blocks (or insert after a block). Best for text, headings, lists, simple tables — no block IDs needed. - update_block: Update content of an existing text, code, or list_item block in-place. - create_block: Create a new block at a precise position. Use parent_block_id to nest inside notice_box, table cell, or layout cell. - delete_block: Remove any block. The ONLY option for BOARD, WIDGET, DOC embed, and GIPHY blocks. - replace_block: Delete a block and create a new one in its place (use when update_block is not supported). - add_comment: Create a new comment or reply on the document (doc-level, block-level, or text-selection). WHEN TO USE EACH OPERATION: - text / code / list_item → update_block. Use replace_block to change subtype (e.g. NORMAL_TEXT→LARGE_TITLE) - divider / table / image / video / notice_box / layout → replace_block (properties immutable after creation) - BOARD / WIDGET / DOC / GIPHY → delete_block only GETTING BLOCK IDs: Call read_docs with include_blocks: true — returns id, type, position, and content per block. BLOCK CONTENT (delta_format): Array of insert ops. Last op MUST be {insert: {text: "\n"}}. - Plain: [{insert: {text: "Hello"}}, {insert: {text: "\n"}}] - Bold: [{insert: {text: "Hi"}, attributes: {bold: true}}, {insert: {text: "\n"}}] - Mention user/doc/board: [{insert: {text: "Hey "}}, {insert: {mention: {id: 12345, type: "USER"}}}, {insert: {text: "\n"}}] — type is USER, DOC, or BOARD. id is numeric (user IDs from list_users_and_teams) - Inline column value: [{insert: {column_value: {item_id: 111, column_id: "status"}}}, {insert: {text: "\n"}}] - Supported attributes: bold, italic, underline, strike, code, link, color, background (not applicable to mention/column_value ops) IMAGE WITH ASSET: For asset-based images, use create_block with block_type "image" and asset_id (instead of public_url). add_markdown_content does NOT support asset images — for mixed content, alternate add_markdown_content (text) and create_block (image) operations in sequence. COMMENTS: - add_comment: Create a new comment or reply on the document. Three scopes: - Doc-level (no block_id): comment appears on the doc as a whole. - Block-level (block_id only): comment is anchored to a specific block. The block shows a comment indicator in the UI. - Text-selection (block_id + selection_from + selection_length): comment is anchored to a specific character range inside a text/code/list_item block. That text is highlighted with a comment marker. Block-level and text-selection comments only work on blocks with text content (text, code, list_item, title, quote). They do NOT work on: divider, page_break, table, layout, notice_box, image, video, or giphy blocks. Get block IDs from read_docs with include_blocks: true. Format body with HTML, not markdown. Use mentions_list for @mentions.

Execute read-only GraphQL queries against the monday.com API. Only queries are accepted : mutations are rejected with an error before the request is sent. Use get_graphql_schema and get_type_details tools first to understand the schema before crafting your query.

Create a monday.com form. Also creates a backing board to store responses. Returns the formToken for future mutations.

Create a new item with provided values, create a subitem under a parent item, or duplicate an existing item and update it with new values. Use parentItemId when creating a subitem under an existing item. Use duplicateFromItemId when copying an existing item with modifications.[REQUIRED PRECONDITION]: Before using this tool, if new columns were added to the board or if you are not familiar with the board's structure (column IDs, column types, status labels, etc.), first use get_board_info to understand the board metadata. This is essential for constructing proper column values and knowing which columns are available.

Create a new board view (tab) with optional filters and sorting. This creates a saved view on a monday.com board that users can switch to. Filter operators: any_of, not_any_of, is_empty, is_not_empty, greater_than, lower_than, between, contains_text, not_contains_text Example filter for people column: { "rules": [{ "column_id": "people", "compare_value": ["person-12345"], "operator": "any_of" }] } Example filter for status column: { "rules": [{ "column_id": "status", "compare_value": [1], "operator": "any_of" }] }

Get updates (comments/posts) from a monday.com item or board. Specify objectId and objectType (Item or Board) to retrieve updates. For Board queries, you can filter by date range using fromDate and toDate (both required together, ISO8601 format). By default, Board queries return only board discussion. Set includeItemUpdates to true to also include updates on individual items. Returns update text, creator info, timestamps, and optionally replies and assets.

Move a folder, board, or overview in monday.com. Use position for relative placement based on another object, parentFolderId for folder changes, workspaceId for workspace moves, and accountProductId for account product changes.

[UI COMPONENT] Renders an interactive smart assignment interface visualization that the user can see and interact with. IMPORTANT: This is a UI DISPLAY tool - use it to RENDER visual components for the user to see and interact with. Do NOT use data-fetching tools when the user explicitly asks to "show", "display", "visualize", or "see" something visually. Helps assign tasks to the right people. Assignment suggestions are based on task details (like name) and person details (such as title, availability, etc).Use for requests to see or use an interactive assignment interface. Always show as much as data possible, while showing the person details like title etc. If you do not have the data available - use the list_users_and_teams tool.

Update a monday.com form. Use the action field to specify the operation.

Update an existing board view (tab) — change its name, filter rules, or sort order. Provide only the fields you want to change. Omitted fields are left unchanged. Filter operators: any_of, not_any_of, is_empty, is_not_empty, greater_than, lower_than, between, contains_text, not_contains_text Example filter for people column: { "rules": [{ "column_id": "people", "compare_value": ["person-12345"], "operator": "any_of" }] } Example filter for status column: { "rules": [{ "column_id": "status", "compare_value": [1], "operator": "any_of" }] }

Creates a new Vibe app from a natural-language prompt. Returns immediately with app_id and editor_link : the URL of the Vibe builder/chat page for the new app (https://{accountSlug}.monday.com/vibe/app/{appId}); the user can open it right away to watch generation in progress. Generation itself runs asynchronously : poll vibe_get for status. Optional: board_ids to connect existing boards (omit to auto-create), view_id to host a dashboard widget, and model to pick the LLM.

Delete a Vibe app and its associated assets. Destructive.

Sends a follow-up message to modify an existing app. Fire-and-forget : returns immediately with user_message_id and editor_link (the Vibe builder/chat URL for this app, https://{accountSlug}.monday.com/vibe/app/{appId}). Returns APP_BUSY (409) if the app is currently generating; poll vibe_get first. Optional model to pick the LLM for this build.

Execute GraphQL mutations against the monday.com API to create, update, or delete data. Only mutations are accepted : queries are rejected with an error before the request is sent. Use get_graphql_schema and get_type_details tools first to understand the schema before crafting your mutation.

Create a monday.com board

Create a new group in a monday.com board. Groups are sections that organize related items. Use when users want to add structure, categorize items, or create workflow phases. Groups can be positioned relative to existing groups and assigned predefined colors. Items will always be created in the top group and so the top group should be the most relevant one for new item creation

Create up to 20 new items in a single call. Each item is fully independent - it chooses its own groupId, parentItemId (for subitems), duplicateFromItemId (for bulk templating from an existing item), and createLabelsIfMissing. A single call can therefore span multiple groups, mix regular items with subitems under different parents, and mix fresh creates with duplicates of existing items. Each item returns its own item_id and item_url on success, or a raw error message on failure. [REQUIRED PRECONDITION]: Before using this tool, if new columns were added to the board or if you are not familiar with the board's structure (column IDs, column types, status labels, etc.), first use get_board_info to understand the board metadata. This is essential for constructing proper column values and knowing which columns are available.

Run arbitrary code in a monday-authenticated sandbox, without saving. Prefer dedicated monday tools for individual reads, writes, and GraphQL queries/mutations : they render in the UI and are retried one step at a time. Reach for execute_code when code is genuinely the better tool: - Bulk / multi-item work : batch operations, dedup, aggregations, joins across boards (one script beats N tool calls that accumulate context and compound failure) - Data transformation : normalizing phones/dates, fuzzy matching, weighted scoring - File I/O : parsing uploaded CSV/XLSX to import items, producing downloadable exports - Multi-step workflows where each step's output gates the next The sandbox has authenticated access to the monday.com API. You can make HTTP requests with GraphQL queries and mutations : authentication is handled automatically. IMPORTANT: Network access is restricted to the following hosts: [api.monday.com, mcp.monday.com/mcp]. Requests to any other host (or a different path on a restricted host) will be blocked. Use this tool to query boards, items, columns, users, updates, and any other monday.com API resource. THE SANDBOX IS PER-CALL: a new empty container every call, destroyed when the call returns. Nothing written to disk survives, /tmp included. Never write a file in one call to read it in a later one, and don't invent staging paths for earlier tool results : none exist. To carry data forward, print it and pass it into the next call's code, or do the whole job in one call. Splitting a fan-out across calls only works if the later calls don't depend on the earlier ones' files. FAIL WITH A NON-ZERO EXIT. A run that prints an error and exits 0 is recorded as a success. The monday.com API returns HTTP 200 with an "errors" array, so check the parsed body rather than the status code and raise when it is present. Let exceptions propagate; don't wrap the script in a bare try/except. TIME LIMIT: 300s. A run that exceeds it is killed, so scope each call to finish well inside the limit instead of fetching everything in one script. Don't call mcp.monday.com from inside the sandbox to reach monday tools : you already have them, and a tool missing from your tool list won't be found there either. Example : monday.com GraphQL query, raising on errors (Python). Use this shape for every API call: code: "import requests\ndef gql(query):\n body = requests.post('https://api.monday.com/v2', json={'query': query}).json()\n if 'errors' in body:\n raise RuntimeError(body['errors'])\n return body['data']\nprint(gql('{ users(limit:5) { id name email } }'))" Example : monday.com GraphQL mutation (Python): code: "import requests\nmutation = 'mutation { create_board(board_name: \"New Board\", board_kind: public) { id } }'\nbody = requests.post('https://api.monday.com/v2', json={'query': mutation}).json()\nif 'errors' in body:\n raise RuntimeError(body['errors'])\nprint(body['data'])" Example : with vars (accessed via os.environ): code: "import os, requests\nuser_id = os.environ['user_id']\nresp = requests.post('https://api.monday.com/v2', json={'query': f'{{ users(ids: [{user_id}]) {{ id name email }} }}'})\nprint(resp.json())" vars: {"user_id": 12345} Example : simple (Python): code: "print('hello world')" Example : with files (input and output): code: "import json\ndata = json.load(open('/tmp/data.json'))\njson.dump({'count': len(data)}, open('/tmp/result.json', 'w'))" files: [{"path": "/tmp/data.json", "content": "W3siaWQiOiAxfV0="}] output_files: ["/tmp/result.json"] Example : auto-collect outputs (write anything you want returned under /outputs): code: "import os, json\nos.makedirs('/outputs', exist_ok=True)\njson.dump({'ok': True}, open('/outputs/result.json', 'w'))" return_outputs: true