Redmine MCP server for Cursor
Enables interacting with a Redmine instance via REST API, allowing issues, projects, time logging, and generic endpoint calls through an MCP agent.
README
Redmine MCP server for Cursor
A small Model Context Protocol (MCP) server that talks to your Redmine instance over its REST API. After you wire it into Cursor, the Agent can list/create/update issues, query projects, log time, and call any documented Redmine endpoint via a generic tool.
This repo runs over stdio (standard MCP transport): Cursor starts node index.js and exchanges JSON-RPC over stdin/stdout. You do not need to expose an HTTP port.
Prerequisites
- Node.js 18+ (includes
fetch) - A Redmine account with an API access key
- In Redmine: My account → API access key (or your admin’s equivalent)
- Permission to use Redmine’s REST API for the actions you care about (varies by role/project)
Quick setup (for anyone cloning this repo)
1. Install dependencies
cd /path/to/this-repo
npm install
2. Configure environment variables
Create a .env file in the same directory as index.js (the repo root). Do not commit .env to git.
RedmineURL=https://your-redmine.example.com
RedmineAPIKEY=your_api_key_here
Rules:
RedmineURL— Base URL only, no trailing slash, e.g.https://redmine.company.comRedmineAPIKEY— Your personal API key (treat it like a password)
Optional check from the shell (does not print your key):
node -e "require('dotenv').config(); console.log(process.env.RedmineURL ? 'RedmineURL OK' : 'missing RedmineURL')"
3. Smoke test (optional)
The MCP entrypoint is meant to be launched by Cursor. If you run it manually it will wait on stdio:
npm start
You should see no errors on startup if .env is valid. Stop with Ctrl+C.
Connecting Cursor
Option A — Project-only MCP (recommended for teams)
Check in .cursor/mcp.json at the repo root so everyone who opens this project gets the same server definition.
This repo’s .cursor/mcp.json uses ${workspaceFolder} so paths work for any clone location:
{
"mcpServers": {
"redmine-all-in-one": {
"command": "node",
"args": ["${workspaceFolder}/index.js"],
"cwd": "${workspaceFolder}"
}
}
}
Why cwd: dotenv loads .env from the current working directory. cwd must be the folder that contains .env and index.js.
If your Cursor build does not expand ${workspaceFolder}, switch to absolute paths under args and cwd for your machine.
Option B — User-global MCP
Add the same server block under your user MCP settings (Cursor’s global MCP config). Use absolute paths to wherever you cloned this repo.
After editing MCP config:
- Open Cursor Settings → MCP (or Features → MCP)
- Confirm
redmine-all-in-oneappears and is enabled - Use Reload / restart Cursor if the server stays disconnected
Using it from chat
- Use Agent mode (or any chat mode where tools are allowed).
- Ask clearly that answers should come from Redmine, e.g.
- “Using Redmine MCP, list my open issues assigned to me.”
- “Create a Feature in project
identifierwith subject…for version….”
- If the model forgets to use tools, say: “Only use Redmine MCP tools for Redmine data.”
The server authenticates with X-Redmine-API-Key; Redmine treats requests as the user who owns that key.
Tools reference
| Tool | Purpose |
|---|---|
redmine_api_request |
Generic REST call: method, path (e.g. /issues.json), optional query string, optional JSON body string. Use this whenever no helper fits. |
redmine_list_projects |
Paginated project list. |
redmine_list_issues |
Issues with common filters (project_id, status_id, assigned_to_id, pagination). Special values often include me, open, *. |
redmine_get_issue |
One issue by id; optional include (e.g. journals). |
redmine_create_issue |
Create issue (project, subject, tracker, dates, assignee, fixed_version_id, estimated_hours, etc.). |
redmine_update_issue |
Partial update; supports estimated_hours. Clear assignee via API: put empty assigned_to_id (see Redmine REST docs / use redmine_api_request). |
redmine_log_time |
POST a time entry (issue_id, hours, optional activity_id, spent_on, comments). Defaults activity_id to 164 if omitted (your server may differ—override when needed). |
redmine_list_users |
List users (permissions required). |
redmine_get_current_user |
Who the API key belongs to. |
For endpoint paths and payloads, your installation follows Redmine’s REST API (see Redmine REST API).
Practical notes (save debugging time)
Validation errors (HTTP 422)
Your Redmine may enforce:
- Priority required on create → set
priority_id(discover viaGET /enumerations/issue_priorities.json). - Estimated time on certain trackers → set
estimated_hourson create/update or inextra_fields_jsonif using an older MCP schema cache.
Clearing assignee
Some setups accept PUT /issues/:id.json with "assigned_to_id": "". If your Redmine behaves differently, use the UI or check your Redmine version’s REST notes.
Security
- Rotate the API key if it was ever committed or pasted in an insecure channel.
- Prefer per-user keys—not shared accounts.
- Restrict file permissions on
.envon shared machines (chmod 600 .env).
MCP schema cache
After upgrading index.js, reload MCP servers in Cursor so new/changed tool arguments appear in the UI.
Development
- Entry:
index.js - Start:
npm start→ runsnode index.js - Dependencies:
@modelcontextprotocol/sdk,dotenv,zod
License
MIT (see LICENSE). Dependencies retain their own licenses; Redmine is a separate project under its own license.
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。
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。
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。
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。