ClickUp MCP Server
Enables Claude Code and Claude Desktop to manage ClickUp tasks, sprints, reports, workspace navigation, and docs through 25+ MCP tools.
README
ClickUp MCP Server
A Model Context Protocol (MCP) server that connects Claude Code and Claude Desktop to ClickUp. Provides task management, sprint tracking, reporting, workspace navigation, and Doc creation through 25 tools.
Features
- Task management — create, update, search, and bulk-edit tasks
- Sprint tracking — auto-detect current sprint, list sprint tasks, filter by assignee/status
- Reporting — sprint reports with at-risk detection, PR link extraction, status summaries
- Workspace navigation — browse spaces, folders, lists; resolve custom task IDs
- Comments — read and write task comments
- Docs — create a native ClickUp Doc with real markdown tables/headers, unlike a task description
Prerequisites
- Python 3.12+
- uv (
brew install uv) - A ClickUp personal API token
Quick Start
uv tool install "clickup-mcp-server @ git+https://github.com/asaphe/clickup-mcp.git"
clickup-mcp-server setup --code
The setup script walks you through:
- Token retrieval (1Password integration or manual paste)
- Workspace configuration (workspace ID required, sprint/team IDs optional)
- Client registration (Claude Code, Claude Desktop, or both)
- Automatic restart of Claude Desktop if running
To configure Claude Desktop instead, run:
clickup-mcp-server setup --desktop
To configure both clients:
clickup-mcp-server setup --both
Alternative: Run from a Local Clone
Use this flow when you want to hack on the server itself:
git clone https://github.com/asaphe/clickup-mcp.git
cd clickup-mcp
uv sync --dev
uv tool install --editable .
python3 setup_mcp.py
This uses the same setup wizard, but runs it from the repo root layout used by this public package.
Run uv tool install --editable . first: the setup wizard only installs
clickup-mcp-server via uv tool install when it isn't already on your PATH,
so an editable install done beforehand is preserved instead of being replaced
by the published GitHub version.
Configuration
All configuration is via environment variables. Only CLICKUP_API_TOKEN and WORKSPACE_ID are required — the rest enable optional features.
Required
| Variable | Description |
|---|---|
CLICKUP_API_TOKEN |
Personal API token from https://app.clickup.com/settings/apps |
WORKSPACE_ID |
Your ClickUp workspace (team) ID |
Optional — Sprint Detection
| Variable | Description |
|---|---|
DEVELOPMENT_SPACE_ID |
Space ID containing your sprints. Required for sprint tools. |
SPRINTS_FOLDER_ID |
Folder ID within the space that holds sprint lists. Required for sprint tools. |
Optional — Team Labels
| Variable | Description |
|---|---|
COMPONENT_TEAM_FIELD_ID |
Custom field ID for Component/Team labels |
CLICKUP_TEAM_LABELS |
JSON mapping of team names to label IDs (see below) |
Example CLICKUP_TEAM_LABELS:
{"backend": "uuid-1", "frontend": "uuid-2", "example-team": "uuid-3"}
Optional — Tuning
| Variable | Default | Description |
|---|---|---|
API_BASE_URL |
https://api.clickup.com/api/v2 |
ClickUp API base URL |
REQUEST_TIMEOUT |
15.0 |
HTTP request timeout in seconds |
MAX_RETRIES |
3 |
Max retry attempts for failed/rate-limited requests |
Finding Your IDs
- Workspace ID: Settings → Workspaces → look at the URL or API response
- Space ID: Click on a Space → the ID is in the URL (
/s/{space_id}/...) - Folder ID: Click on a Folder → inspect the URL or use the
get_workspace_hierarchytool - Custom Field ID: Use the ClickUp API:
GET /list/{list_id}/field - Team Label IDs: Use the ClickUp API:
GET /list/{list_id}/field→ find the labels dropdown field → extract option IDs
Tools
28 tools across 6 categories:
Sprint Management
| Tool | Description |
|---|---|
get_current_sprint |
Auto-detect the active sprint by date |
refresh_sprint_cache |
Force-refresh the cached sprint |
get_sprint_tasks |
List tasks in the current sprint (filter by assignee/status) |
Task Management
| Tool | Description |
|---|---|
get_task |
Get task details by ID (DEV-1234 or UUID) |
create_task |
Create a task in any list |
create_sprint_task |
Create a task in the current sprint (auto-assign, team, points) |
update_task |
Update status, assignee, description, points, dates |
search_tasks |
Search tasks by name |
get_my_tasks |
Get tasks assigned to the current user |
get_list_tasks |
Get all tasks in a specific list |
move_task_to_list |
Move a task to a different list |
bulk_update_tasks |
Batch-update status/team/assignee across multiple tasks |
ensure_task_fields |
Check and fix missing fields (assignee, team, points) |
delete_task |
Delete a task permanently |
Comments
| Tool | Description |
|---|---|
add_task_comment |
Post a comment on a task |
get_task_comments |
Retrieve comments from a task |
Reporting
| Tool | Description |
|---|---|
get_sprint_report |
Sprint report by assignee with status counts and at-risk flags |
Workspace
| Tool | Description |
|---|---|
get_current_user |
Show the authenticated user |
get_workspace_hierarchy |
Browse spaces, folders, and lists |
task_url |
Get the ClickUp URL for a task |
add_tag_to_task |
Add a tag to a task |
list_teams |
List available team labels |
check_team_labels |
Check configured team labels against live ClickUp options |
Docs
| Tool | Description |
|---|---|
create_doc |
Create a native ClickUp Doc (v3 API) with one page of markdown content; returns its URL. Create-only, PRIVATE by default — sharing is a manual follow-up in the ClickUp UI. |
update_doc_page |
Overwrite an existing Doc page's content in place (replace-only), with an optional retitle; returns its URL. |
get_doc |
Get a Doc's metadata (name, parent location, visibility). |
get_doc_pages |
Get every page of a Doc, content included, as a flat list — sub-pages are flattened in, not dropped. |
get_doc_page |
Get a single page's content by doc_id/page_id. |
ClickUp's v3 Docs API has no DELETE endpoint for Docs or Pages — confirmed against
developer.clickup.com's own llms.txt index and public OpenAPI spec, and empirically
(a live DELETE call against a real doc returned 405 Method Not Allowed, not 404,
meaning the route exists but the verb is rejected). So there's no delete_doc tool, and
update_doc_page is replace-only by design: the client retries transport errors,
including a timeout that can fire after the server already applied the edit — a replace
is idempotent under retry, but append/prepend would silently double-apply.
ClickUp's public API also has no Doc/page comment endpoint (only task/list/view comments exist) — reading a Doc's comment panel still requires the browser.
Usage Examples
In Claude Code or Claude Desktop, just ask naturally:
- "show my tasks"
- "sprint report for backend"
- "create a task for fixing the login bug"
- "mark DEV-1234 as done"
- "what's the current sprint?"
- "show unassigned tasks in the sprint"
- "write this up as a doc so the tables render properly"
- "read that doc back to me"
Key Patterns
- Task IDs: Both custom IDs (
DEV-1234,PROJ-456) and UUIDs are accepted everywhere - "me" as assignee: Resolved automatically from the API token
- Sprint auto-detection: Current sprint is detected by date and cached per session
- Rate limiting: Automatic retry with exponential backoff on 429 responses
- Concurrent comment fetching: Sprint reports fetch PR links from comments with bounded concurrency
Coexistence with Built-in ClickUp MCP
This server works alongside the official ClickUp MCP connector. The built-in handles features not covered here (time tracking, chat, and a Doc's comment panel — ClickUp's public API has no Doc-comment endpoint). Doc create/read/update are covered here; deletion isn't possible via either server, since the v3 Docs API exposes no delete endpoint. You can register both, though having both may cause tool-selection ambiguity for overlapping operations.
Development
uv sync --dev
uv run pytest
uv run mypy clickup_mcp_server
uv run ruff check .
Architecture
clickup_mcp_server/
server.py — FastMCP server entry point and instructions
client.py — Async HTTP client with retry and rate-limit handling
config.py — Settings (env var based, no hardcoded IDs)
models.py — Pydantic models for API responses
tools/
sprint.py — Sprint detection and caching
tasks.py — Task CRUD, search, bulk operations
comments.py — Comment read/write
reporting.py — Sprint reports with at-risk detection
workspace.py — User info, hierarchy, tags
docs.py — Doc create/read/update (v3 API)
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。