ClickUp MCP With Claude Code: Setup and Limits

To use ClickUp from Claude Code, add ClickUp’s hosted MCP server with claude mcp add --transport http clickup https://mcp.clickup.com/mcp, then run /mcp inside a Claude Code session and sign in to ClickUp in the browser. From then on Claude Code can search your Workspace, read and create tasks, update statuses and assignees, comment, log time and work with Docs, all with your own ClickUp permissions. Sign-in is OAuth only; ClickUp does not accept API keys on this server. The catch most people meet first is the rate limit: without ClickUp’s AI add-on, MCP calls are capped per rolling 24 hours. The rest of this guide covers each step and each limit.

ClickUp describes the server↗ (opens in a new tab) as being in public beta and available on all of its plans. It is a remote server that ClickUp runs, so there is nothing to install. If MCP itself is new to you, start with what MCP is.

Before you start

• A ClickUp account with access to the Workspace you want to use.
• Claude Code installed and signed in.
• A browser on the same machine for the sign-in. If you work over SSH, see the section on running without a browser below.
• An idea of your Workspace plan and whether it has the Everything AI add-on, because that decides your daily call allowance.

Step 1: add the server

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

This is the command in ClickUp’s setup instructions↗ (opens in a new tab). clickup is only the local name. --transport http selects streamable HTTP, which is what the server speaks. By default Claude Code stores the server in local scope, for you and this project only, in ~/.claude.json. Add --scope user to have it in every project on your machine, or --scope project to write it to .mcp.json for the team to commit. With project scope nobody shares a login: the file holds only the address, each person signs in with their own ClickUp account, and Claude Code asks before it uses a server it found in that file.

Step 2: sign in

1. Start Claude Code in your project and type /mcp.
2. Pick the ClickUp server, which will show as needing authentication, and choose to authenticate.
3. Your browser opens on ClickUp. Sign in and authorise the Workspace you want Claude Code to work in. Check the app name and the Workspace before approving.
4. Back in Claude Code, /mcp should show the server connected, with its tools listed.

Claude Code stores the token and refreshes it for you, according to its documentation↗ (opens in a new tab); you never see or paste it. What happens during that browser step, and what to look for on the approval screen, is in how MCP sign-in works.

Step 3: first prompts to try

Start with reads. Every call counts against the daily allowance, so a few precise questions teach you more than a broad one.

• “Show me the Spaces, Folders and Lists in this Workspace.” This uses the hierarchy tool and tells you what names to use in later prompts.
• “Find the tasks tagged bug in the Sprint 3 List that are assigned to me.”
• “Summarise the comments on the ‘Checkout redesign’ task and list the open questions.”
• “How long has ‘Payment retries’ been In Progress?” ClickUp has time-in-status tools for exactly this.

Then try one write you can check by eye: “Create a task ‘Retry job double-charges on timeout’ in the Bugs List, assign it to me and add a comment with the steps to reproduce.” ClickUp’s tool list↗ (opens in a new tab) also covers bulk creation and bulk updates, task links and dependencies, moving tasks between Lists, time entries and timers, Docs pages and posting to Chat channels.

Permissions: what it can and cannot do

ClickUp’s documentation is plain about the ceiling: the assistant can only do what you have permission to do in ClickUp. There is no separate, narrower identity for it. Whatever your account can change in the authorised Workspace, Claude Code can change too, through the tools the server exposes.

On deletion, ClickUp’s pages disagree with each other. Its FAQ says deletion tools have not been added, as a safety measure, while its tools page lists a “Delete task” tool. Treat the tool list that /mcp shows you as the truth for your connection, and do not rely on deletion being impossible.

The layer you control is Claude Code’s own permission prompts. In Manual mode it asks before calling an MCP tool (recent versions start in auto mode instead), and you can pre-approve individual tools in settings with rules↗ (opens in a new tab) of the form mcp__clickup__<tool>, using the exact tool names /mcp shows. Pre-approve searches and reads. Keep anything that creates, updates, moves or posts on ask until you trust your prompts, and leave any delete tool on ask for good.

{
  "permissions": {
    "allow": [
      "mcp__clickup__<search tool>",
      "mcp__clickup__<get task tool>"
    ]
  }
}

Running it without a browser

Because this server takes OAuth only, there is no token to paste for a script or a CI runner. On a machine without a display, sign in with claude mcp login clickup --no-browser: Claude Code prints the sign-in address for you to open elsewhere, as its documentation describes. For clients that cannot speak remote MCP directly, ClickUp documents the mcp-remote bridge (npx -y mcp-remote https://mcp.clickup.com/mcp), which still signs in through the browser.

Limits to know

• Daily call limits. Without the Everything AI add-on, ClickUp documents 50 MCP calls per rolling 24 hours on the Free Forever plan and 300 on Unlimited and above. The window starts at your first request, cannot be reset, and calls made while limited do not extend it. With the add-on, MCP limits match ClickUp’s public API limits for your plan. ClickUp says these limits may change.
• One prompt can be several calls. A request such as “move every Ready for Review task to In Review” may search, read and update, and each call counts.
• OAuth only. No personal API keys or your own access tokens on this server.
• Public beta. ClickUp says the server is in public beta and that further restrictions may apply.
• No Connected Search. ClickUp’s FAQ says the MCP server cannot search apps connected to ClickUp.
• Large results fill context. Claude Code warns when one MCP result passes 10,000 tokens and caps output at 25,000 by default (MAX_MCP_OUTPUT_TOKENS raises it). A whole Workspace hierarchy can be big; ask about one Space at a time.

Using it in Claude on the web or desktop

The same server works as a custom connector in Claude: add a connector with the name ClickUp and the URL https://mcp.clickup.com/mcp, then sign in. Anthropic’s help centre↗ (opens in a new tab) says custom connectors are available on all Claude plans, with the Free plan limited to one, and that on Team and Enterprise an Owner adds the connector for the organisation first. Connectors you add in Claude also appear in Claude Code when you are signed in with the same Claude account, marked as coming from claude.ai under /mcp.

Removing or resetting it

claude mcp logout clickup   # clear this machine's sign-in
claude mcp remove clickup   # remove the server entirely

The “Clear authentication” option under /mcp does the same as logout. That removes the token from your machine; access on the ClickUp side is managed in ClickUp.

If ClickUp is more than you need

ClickUp is a platform, with Docs, goals, time tracking and dashboards beside tasks, and the MCP server exposes much of it. If what you actually want is one board your assistant works on as a member, holding a role on the board rather than all of your rights, with every change it makes recorded in the board’s history under its own name, that is fenbs. The fair comparison is on fenbs vs ClickUp, and the Claude Code setup is on its integration page.