ClickUp MCP Server: Tools, Limits and How to Connect It
What ClickUp's official MCP server can do, how to connect it to Claude, ChatGPT or Cursor, and the daily limits to plan around.

If your team runs its work in ClickUp and you draft in Claude, ChatGPT or Cursor, the gap between them is copy-paste: a task description into the chat, the answer back into ClickUp. ClickUp's official MCP server closes that gap: connect it once and your AI assistant can read and change your Workspace directly, within your own permissions.
That matters most if you search, triage and update tasks all day. Below: setup in each client, the 48 tools, the daily rate limits that catch most people out, and where the server stops.
Is there an official ClickUp MCP server?
Yes. ClickUp MCP is ClickUp's own server that exposes your Workspace to an AI assistant through the Model Context Protocol, and ClickUp hosts it and offers it on all plans (ClickUp Help). Its help center names ChatGPT, Claude, Cursor and VS Code as clients; the developer docs say it supports "all major MCP clients and Apps."
MCP is the open standard Anthropic introduced in November 2024 so AI tools can connect to outside systems through one interface (Anthropic, 2024). The ClickUp MCP server is one server built on that MCP standard. In practice it's a connector: it gives your AI a place to look, not something it already knows. Connectors vs MCP vs a Context Layer explains why that distinction matters.
Official vs community ClickUp MCP servers
Before ClickUp shipped its own, developers built community servers on the ClickUp API. The best known, taazkareem/clickup-mcp-server, says it isn't affiliated with ClickUp, now sells licenses from $9 a month, and runs locally with a CLICKUP_API_KEY and team ID in environment variables. For most people the official server is simpler: no API key in a JSON file, nothing to install, and access you can revoke from ClickUp.
How do you connect ClickUp MCP?
Add the server URL https://mcp.clickup.com/mcp to your MCP client, then sign in to your ClickUp account in the browser and pick a Workspace. Authentication is OAuth only: ClickUp doesn't accept a personal API key or access tokens, and admins don't need to do anything inside ClickUp first. ClickUp's setup guide covers each client.
Claude (claude.ai and Claude Desktop)
- Go to Customize > Connectors, click + Add, then Add custom connector (Claude Help Center).
- Name it "ClickUp" and paste the server URL.
- Click Add, then sign in to ClickUp when Claude asks.
Free plans allow one custom connector. On Team and Enterprise, an owner adds it under Organization settings > Connectors first, and members then click Connect. ClickUp's guide also offers a manual installation for the Claude desktop app: add the server to claude_desktop_config.json through mcp-remote.
Claude Code
- Run the command below.
- Open a session and run
/mcpto finish the OAuth sign-in.
claude mcp add --transport http clickup https://mcp.clickup.com/mcp
ChatGPT
- Open Apps and search for ClickUp, or browse Productivity.
- Click Connect, sign in, select the Workspaces to use and click Connect Workspace.
In ChatGPT Team, Enterprise and Edu workspaces, only owners and admins can configure MCP servers, according to ClickUp.
Cursor
Install ClickUp from the Cursor marketplace, or add it to mcp.json yourself:
{
"mcpServers": {
"clickup": { "url": "https://mcp.clickup.com/mcp" }
}
}
The same guide covers VS Code (one code --add-mcp command) and Microsoft Copilot Studio (add a Model Context Protocol tool with OAuth and dynamic discovery). A custom client connects too, if it speaks JSON-RPC 2.0 over HTTP and supports OAuth 2.1 with PKCE. MCP setup in general is in How to Connect an LLM with MCP.
What can ClickUp MCP do?
It exposes 48 tools, according to ClickUp's tool list. Sorted by name, about 20 read and 28 write.
Read:
- Search: search tasks, Docs and comments across the Workspace, or tasks by type or tag. This is the strongest use: ask in natural language instead of clicking through Lists and filters.
- Tasks and comments: get a task with its task properties, task comments and threaded replies.
- Workspace hierarchy: Spaces, Folders and Lists, plus members and assignees.
- Time: time entries, the current timer, and time in status.
- Docs and Chat: list and read document pages, list Chat channels.
Write:
- Task creation and updates: create a new task, update existing tasks, create or update multiple tasks in bulk, set a custom field value, move tasks between Lists, attach files, add tags, links and dependencies.
- Comments and Chat: post a comment or a Chat message.
- Time tracking: start and stop timers, add entries.
- Docs and hierarchy: create Docs and pages, update pages, create or update Lists and a new Folder.
ClickUp's examples are the requests you'd expect: "Create a task for this bug we're working on," or "Find all my incomplete tasks due this week."
The gaps matter if you plan task workflows around it. There are no tools for Goals or webhooks, and none creates new custom fields. The only deletion tool is Delete task; Lists, Folders, Docs and comments can't be deleted through MCP. Search stays inside your Workspace, so Connected Search won't reach apps linked to ClickUp. The tool list changes, so check the supported tools before you build on one.
What are the ClickUp MCP rate limits?
Without the Everything AI add-on, ClickUp caps MCP calls per rolling 24 hours by plan (ClickUp developer docs):
| ClickUp plan | MCP calls per 24 hours |
|---|---|
| Free Forever | 100 |
| Core / Unlimited | 300 |
| Business | 1,000 |
| Business Plus | 2,500 |
| Enterprise | 5,000 |
| Enterprise Plus | 25,000 |
Every tool call counts once, reads included; the initialize and tools/list handshakes don't. The window starts with your first request and can't be reset, and calls over the limit return a 429 RATE_LIMIT_EXCEEDED error with a retryAfter value. You can't see how many calls your Workspace has used. The help center, updated 23 October 2026, lists the same numbers up to Enterprise and adds that ClickUp's newer plans with AI included get different daily limits.
With the Everything AI add-on, the daily cap goes away and MCP shares the public ClickUp API's per-minute limit: 100 requests per minute per token on Free Forever, Unlimited and Business, 1,000 on Business Plus and 10,000 on Enterprise (ClickUp Help).
Do Claude and Cursor share the same limit?
ClickUp's own pages disagree. The developer docs say limits are "per Workspace and per client," so Claude and Cursor would each get 300 on Unlimited. The help center says every MCP client connected to the same Workspace draws from the same pool. Plan for the shared pool.
Plan for the multiplier too. One question like "what's overdue for my team this week?" can take a search, a few task reads and a member lookup, so 100 calls on the Free Forever plan runs out faster than it sounds.
ClickUp MCP or the ClickUp API: which should you use?
Use MCP when a person is asking questions in an AI assistant, and the ClickUp API when you're building something that runs on its own. The API takes a personal API key, offers webhooks and covers objects MCP doesn't, such as Goals. MCP needs no code and keeps access tied to your OAuth login. For custom agents that run unattended at volume, the daily MCP caps decide it.
Is ClickUp MCP safe to connect?
It's as safe as the access your ClickUp account has, with one real risk: the AI reads task descriptions and comments that other people wrote, and any of them can carry a prompt injection. In the MCPTox benchmark, poisoned tool descriptions on 45 real MCP servers got o1-mini to follow them 72.8% of the time (Wang et al., 2025).
As a safety measure, keep write tools on approval rather than "always allow," turn off tools you don't need, and don't pair ClickUp MCP in one session with a server that reads public content. OAuth limits the connection to what your account sees, and you can revoke it. The full threat model is in Is MCP Safe?
Does ClickUp MCP cost you context tokens?
It can, depending on your client. 48 tool definitions take room in the context window that your actual work could use (why too many MCP servers make your AI worse). Claude Code turns on tool search by default, so only tool names and server instructions load at session start (Claude Code docs). In clients that load every definition up front, connect ClickUp for the sessions where you'll use it.
Is ClickUp MCP worth connecting?
Yes, if you want to search and edit your Workspace inside a live AI session. Asking for overdue tasks, turning a meeting summary into tasks with due dates, or pulling a Doc in as reference is faster than switching tabs.
It's the wrong tool for event-driven work, which still belongs in ClickUp Automations or webhooks, and for high-volume jobs on the lower plans. And it only knows ClickUp: it can tell your AI what's in your Workspace, not what was decided in last week's email thread.
Where does Unabyss fit next to ClickUp MCP?
Unabyss covers the part a single-app server doesn't: background from all your work apps, not just ClickUp. It imports your ClickUp data when you connect and re-imports it on a schedule, alongside Gmail, Google Calendar, Slack, Notion, GitHub and the other apps on its integrations page, and serves that context to Claude, ChatGPT and Cursor over one MCP connection.
Unabyss doesn't act inside ClickUp: it won't create, update or close a task. Many people run both, ClickUp MCP for actions and Unabyss for context. In an August 2026 Unabyss benchmark, adding that context to the same connected tools cut cost by 36% and follow-up prompts by 60% (What Structured AI Context Is Worth). The other vendor servers are compared in Best MCP Servers for Work. See how Unabyss works.