MCP Server for Redmine

Model Context Protocol (MCP) server for Redmine project management integration. Allows AI assistants to create, update, and query Redmine issues directly from conversations.

Features

• ✅ Create issues with full metadata (priority, estimates, parent tasks)

• ✅ Update existing issues (status, description, comments)

• ✅ List and filter issues by project and status

• ✅ Get detailed issue information

• ✅ List all accessible projects

• ✅ Full wiki access — list every page, read, create and update from a file for large content, or apply a targeted append/prepend/replace patch that leaves the rest of the page untouched

Related MCP server: Redmine MCP Server

Installation

cd mcp-server-redmine
npm install
npm run build

Configuration

Set environment variables for your Redmine instance:

export REDMINE_URL="https://your-redmine.com"
export REDMINE_API_KEY="your-api-key-here"

Getting Your Redmine API Key

IMPORTANT: You need the REST API key, not the Atom feed key!

If REST API is Enabled:

1. Log into Redmine

2. Go to "My account" (top right)

3. Look for "API access key" section (right sidebar)

4. Click "Show" to reveal the key

5. Copy the 40-character key

If You Only See "Atom Access Key":

The Atom key is for RSS feeds only. You need REST API enabled:

1. Admin must enable it:

   • Administration → Settings → API

   • Check "Enable REST web service"

   • Save

2. Then get your key:

   • Refresh "My account" page

   • "API access key" section should now appear

   • Click "Show" and copy

3. Alternative: Try using Atom key

   • Some Redmine versions accept it

   • Test with npm test

   • If 401 error, you need proper API key

Usage with Amazon Q

Add to your Amazon Q configuration file (~/.aws/amazonq/mcp.json or IDE settings):

{
  "mcpServers": {
    "redmine": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-server-redmine/build/index.js"],
      "env": {
        "REDMINE_URL": "https://your-redmine.com",
        "REDMINE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Available Tools

create_issue

Create a new Redmine issue/task.

Parameters:

• project_id (required): Project ID number

• tracker_id: Tracker type (1=Bug, 2=Feature, 3=Support) - defaults to 2 (Feature)

• subject (required): Issue title

• description: Issue description (markdown supported)

• priority_id: Priority level (1=Low, 2=Normal, 3=High, 4=Urgent, 5=Immediate)

• estimated_hours: Time estimate in hours

• parent_issue_id: Parent issue ID for creating subtasks

Example:

Create a Redmine issue in project 1 titled "Implement ServingsService" 
with description "Add quantity scaling logic" and estimate 8 hours

update_issue

Update an existing issue.

Parameters:

• issue_id (required): Issue ID to update

• tracker_id: Change tracker type (1=Bug, 2=Feature, 3=Support)

• subject: New title

• description: New description

• status_id: Status (1=New, 2=In Progress, 3=Resolved, 5=Closed)

• priority_id: Priority level

• notes: Add a comment

Example:

Update Redmine issue #123 status to 3 (Resolved) and add note "Implementation complete"

list_issues

List issues from a project.

Parameters:

• project_id (required): Project ID

• status_id: Filter by status ("open", "closed", or specific ID)

• limit: Max results (default 25)

Example:

List all open issues in Redmine project 1

get_issue

Get detailed information about a specific issue.

Parameters:

• issue_id (required): Issue ID

• include_journals: Set to true to include the comment/note history (author, timestamp, text). Journal entries that only record field changes (status flips, done-ratio edits) are filtered out. Default false to keep the response small.

Example:

Get details of Redmine issue #123 including its comments

list_projects

List all accessible Redmine projects.

Example:

List all my Redmine projects

Wiki tools

list_wiki_pages

List every page in a project's wiki, with version, last-updated date and parent. sort accepts title (default), updated (most recently changed first — the useful order for spotting stale pages) or version (most-edited first).

Use it before any wiki-wide operation. It returns the real page set, including pages that no index links to — a sweep driven off the [[links]] in an index page silently skips orphans, and an orphan is exactly the page most likely to have gone stale. Titles come back as Redmine identifiers, so they can be passed verbatim as title to the other wiki tools.

get_wiki_page

Read a wiki page.

raw: true returns only the page body — no title heading, no version footer. Use it whenever the text will be read back and rewritten: the decorated default is not round-trip safe, because re-submitting it would bake the heading and footer into the page itself.

create_wiki_page / update_wiki_page

Write a page. Supply the body either inline via text or from a file via file_path — exactly one of the two. Supplying both is refused rather than silently picking one, and supplying neither is refused rather than blanking the page.

file_path reads a UTF-8 file from disk and writes its contents verbatim. Prefer it for anything large: nothing has to pass through the model, so a big page can't be truncated or subtly reworded in transit.

⚠️ update_wiki_page replaces the entire page body. It is not a patch. If you only mean to change part of a page, use patch_wiki_page — a full overwrite built from a partial body is the most common way to lose wiki content.

patch_wiki_page

Change part of a page without resending the whole body. Fetches the current text, applies one edit, writes it back — so untouched content never passes through the caller and cannot be lost.

Modes:

Two guards make a bad edit fail instead of succeeding wrongly:

• expect_count (replace mode, default 1) — the write is refused unless find matches exactly that many times. A find matching nothing is a no-op nobody notices; a find matching five times when one was meant is an accidental mass edit. Both become errors. Pass expect_count explicitly to authorise a genuine multi-match replace.

• check_version (default true) — the version read during the fetch is sent back, so Redmine refuses the write if the page changed in between. Set it to false only to deliberately overwrite someone else's concurrent edit.

find is a literal string, not a regex, so $, ., ( and friends need no escaping.

Examples:

Append a new row to the tracking table on the Epic_921 wiki page

Replace "Status: paused" with "Status: active" on Closed_Beta_Focus

Update Project_Status from C:\work\project-status.md

Example Workflows

1. Convert Documentation to Tasks

You: "Read PRODUCT_ROADMAP.md and create Redmine issues for all Phase 1 tasks in project 1"

AI: 
- Reads PRODUCT_ROADMAP.md
- Extracts Phase 1 tasks
- Creates issues with:
  - Subject from task name
  - Description from details
  - Priority based on roadmap priority
  - Estimated hours from effort estimates

2. Update Task Status

You: "Mark Redmine issue #45 as resolved since ServingsService is complete"

AI: Updates issue #45 with status_id=3 and adds completion note

3. Create Subtasks

You: "Create subtasks for Redmine issue #50 to implement Categories feature:
- Database migration
- Service layer
- Controller updates
- UI components"

AI: Creates 4 subtasks with parent_issue_id=50

4. Sprint Planning

You: "List all open issues in project 1 and create a sprint plan"

AI: 
- Lists issues
- Analyzes priorities and estimates
- Suggests grouping and ordering

Common Redmine Status IDs

• 1 = New

• 2 = In Progress

• 3 = Resolved

• 4 = Feedback

• 5 = Closed

• 6 = Rejected

(Your Redmine instance may have different IDs - check your Redmine settings)

Common Redmine Priority IDs

• 1 = Low

• 2 = Normal

• 3 = High

• 4 = Urgent

• 5 = Immediate

Troubleshooting

"I only see Atom Access Key, not API Access Key"

See TROUBLESHOOTING_API_KEY.md for complete solutions.

Quick fix: Ask your admin to enable REST API:

• Administration → Settings → API → "Enable REST web service"

"REDMINE_URL and REDMINE_API_KEY environment variables are required"

Make sure environment variables are set in your MCP configuration.

"401 Unauthorized"

Check that your API key is correct and has not expired.

"403 Forbidden"

Your API key doesn't have permission for the requested operation.

"404 Not Found"

Project or issue ID doesn't exist or you don't have access.

Development

# Watch mode for development
npm run watch

# Build for production
npm run build

License

MIT