monday.com MCP With Claude: Setup Guide

monday.com runs a hosted MCP server at https://mcp.monday.com/mcp, and it comes preinstalled on every monday.com account, on every plan. To use it with Claude, add monday.com from Claude’s connector directory and approve the sign-in. To use it with Claude Code, run claude mcp add --transport http monday https://mcp.monday.com/mcp and sign in with /mcp. Either way Claude acts with your own monday.com permissions: it can read boards and items, create and update them, post updates, and build docs and dashboards. The rest of this guide covers both routes, the tools, the permission model, and the admin switches that can block you.

One thing to know before copying an older guide: monday has deprecated its SSE endpoint↗ (opens in a new tab) at https://mcp.monday.com/sse and says it will not come back. Use the /mcp URL with streamable HTTP everywhere. If MCP itself is new to you, start with what MCP is.

Before you start

• A monday.com account as an admin or member. monday says viewers do not have MCP access, and guests have it only on Enterprise when an admin turns it on.
• MCP switched on for your account. An account admin controls it under Administration, AI governance, AI permissions, in the AI Connectors section. If you cannot connect, that is the first thing to ask about.
• For the Claude app: monday’s help centre lists the connector for Claude Pro, Max, Team and Enterprise plans.
• For Claude Code: Claude Code installed and signed in, and a browser for the one-time sign-in.

Route 1: the connector in the Claude app

1. In Claude, open Customize, then Connectors, and find monday.com in the directory.
2. Select Connect. monday.com’s authorisation page opens; review it and approve.
3. Make sure the connector is switched on for your conversation from the tools menu in the chat.
4. Test it with “List my monday.com boards.”

On Claude Team and Enterprise plans it takes two steps: an owner first enables monday.com under the organisation’s connectors, then each member connects it with their own monday.com account. monday describes this as the way to make sure Claude only reaches data each person can already see.

Route 2: Claude Code

claude mcp add --transport http monday https://mcp.monday.com/mcp

Then start Claude Code, type /mcp, choose monday and sign in to monday.com in the browser. monday’s developer documentation↗ (opens in a new tab) lists Claude Code among the coding tools that take the hosted URL and authenticate through OAuth when prompted, and the server publishes the OAuth metadata that lets Claude Code register itself, so there is no client ID to create. monday is only the local name, but it becomes part of every tool’s name in permission rules, so keep it once chosen.

Two shortcuts and one alternative. If you already connected monday.com in the Claude app and sign in to Claude Code with the same claude.ai account, the connector appears in /mcp without any command. Add --scope user to the command to have monday in every project, or --scope project to share the entry through .mcp.json. The alternative is a personal API token, covered below. What happens in the browser step is explained in how MCP sign-in works.

What the tools can do

monday says the server exposes more than 60 tools, and asks clients to treat tools/list as the current list. The groups that matter most for day-to-day work:

• Boards: get_board_items_page, get_board_info, board_insights, get_board_activity, create_board.
• Items: create_item, change_item_column_values, create_update for posting an update, get_updates for reading them.
• Structure: create_group, create_column, list_workspaces, workspace_info, create_folder, move_object.
• Docs, forms and dashboards: create_doc, read_docs, create_form, create_dashboard, create_widget.
• People and search: get_user_context, list_users_and_teams, create_notification, search.
• monday dev sprints: get_sprints_metadata, get_sprint_summary.
• Raw API access: all_monday_api runs any monday GraphQL query or mutation↗ (opens in a new tab), with get_graphql_schema and get_type_details to explore the schema.

That last group matters for permissions. A tool that can run any mutation can do whatever the API lets your account do, including changes no named tool offers. In some clients monday also renders interactive tables, charts and assignment suggestions; those are display tools and depend on the client.

First prompts to try

• “Who am I in monday.com, and which boards do I use most?”
• “What’s overdue across the launch boards? Group it by owner.”
• “Summarise what changed on the Roadmap board this week.”
• “Turn these meeting notes into items on the Sprint board, with owners and due dates, and show me the list before you create anything.”

monday’s own guidance is to review AI-suggested changes before confirming anything that modifies your workspace. Asking Claude to show the plan first, as in the last prompt, is the simplest way to do that.

Permissions: whose rights does it use?

Yours. monday says each user authorises MCP individually, there is no service account or shared key, and an AI agent acting for someone can never see more than that person can in monday.com, including private boards and item-level permissions. Every change is made as the person who connected.

Admins have three levers. They can switch AI Connectors off entirely, or leave them on and uncheck Public Hosted MCP to block Claude and other MCP clients. They can restrict monday MCP to specific workspaces under Apps, monday MCP, Permissions. And they can turn off the hosted connector and require an OAuth app of their own↗ (opens in a new tab), which lets them cap what connections can do with scopes such as boards:read, even for users with wider rights. That last route is the only way to make a monday connection read-only on the monday side.

On the Claude side, the Claude app lets you set each connector tool↗ (opens in a new tab) to Always allow, Needs approval or Blocked. In Claude Code, pre-approve reads and keep everything else on ask:

{
  "permissions": {
    "allow": [
      "mcp__monday__get_*",
      "mcp__monday__search"
    ],
    "deny": [
      "mcp__monday__all_monday_api"
    ]
  }
}

Denying all_monday_api is optional, but it keeps Claude Code to the named tools, which are easier to review.

Running it without a browser

monday also accepts a personal API token as a Bearer token, which suits scripts and machines where nobody can sign in. You find it in monday.com under your profile picture, Developers, My access tokens. monday notes↗ (opens in a new tab) that a personal token mirrors your own permissions and is meant for individual use, not integrations other people share. Keep it in an environment variable, never in a committed file.

claude mcp add --transport http monday https://mcp.monday.com/mcp --header "Authorization: Bearer $MONDAY_TOKEN"

If you would rather keep OAuth on a remote machine, claude mcp login monday --no-browser prints the sign-in address for you to open elsewhere and asks for the redirect URL back.

Limits to know

• Calls count against your API limit. monday says MCP tool calls consume the account’s daily API call limit, the same as direct API requests, which adds up for long agent sessions.
• Which tools you see depends on your monday permissions, plan features and client.
• Large boards fill context. Claude Code warns when a tool result passes 10,000 tokens and caps it at 25,000 by default. Ask for one group, one owner or a filtered page.
• It acts as you, with no separate identity for the assistant.

Removing it

In the Claude app, disconnect monday.com under Customize, then Connectors. In Claude Code, claude mcp logout monday clears the sign-in and claude mcp remove monday removes the server; if you used a token, revoke it in monday as well. An admin can cut off every connection at once from the AI Connectors settings.

If monday.com is more than you need

monday.com is a configurable work platform, with columns, automations and dashboards you build, and its MCP server can reach nearly all of it with your rights. If what you want instead is a board with four fixed lanes and three kinds of task, where your assistant is a member with a role you chose and its name on every change, that is what fenbs is for. The fair comparison is on fenbs vs monday.com, and the setup is on connect Claude and connect Claude Code.