clickup-custom-mcp
An MCP server for interacting with ClickUp via its REST API, providing tools to manage tasks, search work, and explore workspace structure.
README
clickup-custom-mcp
A stdio MCP server that talks to the ClickUp REST API v2 directly.
It exists because ClickUp's hosted MCP (mcp.clickup.com) caps a Free workspace at
50 calls per 24 hours and refuses personal API tokens, while the same account's REST
API allows 100 requests per minute. This server trades a daily ceiling for a
per-minute one, and keeps the token on your machine instead of in a hosted integration.
Requirements
- Node 20+ (see
enginesinpackage.json) - A ClickUp personal API token
Setup
npm install
npm run build
Generate a token in ClickUp: avatar → Settings → Apps → API Token → Generate.
Copy .env.example to .env at the repo root (.env is gitignored) and fill in the token:
cp .env.example .env
The server loads that file itself via process.loadEnvFile, so the token never has to be
written into Claude Code's config — where claude mcp add -e would store it in plaintext.
Anything already present in the environment wins over .env, and if the file is missing
(or the Node build has no loadEnvFile) the server simply falls back to plain environment
variables.
Environment variables
| Variable | Required | Purpose |
|---|---|---|
CLICKUP_API_TOKEN |
yes | Personal API token. The server exits with a message if it's unset. |
CLICKUP_TEAM_ID |
no | Pin a workspace. Without it the first workspace the token can see is used, and a note goes to stderr if there is more than one. |
CLICKUP_RATE_LIMIT |
no | Requests/minute the client allows itself. Defaults to 90 — deliberate headroom under ClickUp's 100. |
Register with Claude Code
claude mcp add clickup-custom-mcp -- node /absolute/path/to/clickup-custom-mcp/dist/index.js
No -e CLICKUP_API_TOKEN=... is needed — the .env file covers it.
Run it standalone with npm start. stdout is the JSON-RPC channel, so every diagnostic
(including a per-request log line with ClickUp's remaining quota) goes to stderr.
Tools
Six coarse tools, each shaped around a job rather than an endpoint. The fan-out lives in
the server, not in the model: one get_my_work call costs 2 HTTP requests where walking
space → folder → list → task from the model side costs thirty.
| Tool | What it does |
|---|---|
get_workspace_map |
Spaces, folders, lists (with their status names) and members, plus your own numeric user id. Call it first when you need an id for anything else. Cached 15 minutes; optionally includes custom field definitions. |
get_my_work |
Every open task assigned to you across the workspace, grouped by overdue / today / this week / later. The one for a standup or a backlog sweep. |
search_tasks |
Filtered task query across the whole workspace in one request — lists, spaces, assignees, statuses, tags, due/updated date ranges, ordering. Paginates automatically. |
get_task |
Full detail for one task: description, custom field values, subtasks, optionally comments. |
create_task |
Create one or many tasks in a single call. Each is reported individually; one failure does not abort the rest. |
update_task |
Update one or many tasks — status, assignees, due dates, priority, renames, archive. Also reported individually. |
Every tool reports how many HTTP requests it spent, and any result that hit the page cap says Truncated rather than quietly returning a short list.
Design notes
- Rate gate (
src/clickup.ts) — a sliding 60-second window plus a concurrency cap of 4. A fixed-interval refill would let a burst of 90 fire in the first second and then stall for 59; tracking actual send times spreads the same budget. - 429 handling — backoff is driven by ClickUp's own
X-RateLimit-Resetheader, retried up to 3 times. The client also coasts down whenX-RateLimit-Remainingdrops to 2 rather than taking a 429 mid-fan-out. - Structure cache — hierarchy, custom fields and member lists are memoised for 15 minutes. Task data is deliberately never cached.
- Compact output (
src/format.ts) — raw ClickUp task JSON runs 2–4KB each. Every tool projects down to the fields a person actually reads and renders markdown instead of JSON.
Layout
src/index.ts entrypoint: env loading, client + server wiring, stdio transport
src/clickup.ts REST client: rate gate, retry/backoff, TTL cache, task pagination
src/tools.ts the six MCP tool definitions and their fan-out
src/format.ts task projection and markdown rendering
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.