Skip to content
Backlog MCP Server

Backlog MCP Server

by Nulab
Released
v0.18.1
MIT
Repository Docs npm
Productivity

Summary

Nulab's official MCP server for Backlog — let an agent read and write projects, issues, comments, milestones, wiki pages, Git repositories and pull requests through the Backlog API.

Features

  • Full CRUD over Backlog projects, issues, comments, versions/milestones and wiki pages
  • Git repository and pull request tools alongside the issue tracker
  • Field selection to return only the fields a caller needs
  • Token limiting so large result sets degrade instead of overflowing context
  • Exportable, customisable tool descriptions via --export-descriptions
  • Docker image and npx entry point, both configured with two environment variables

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it, filling in the environment variables with your own values.
  3. Restart the agent — the "Backlog MCP Server" tools will be available.
{
  "mcpServers": {
    "backlog-mcp-server": {
      "command": "npx",
      "args": [
        "backlog-mcp-server"
      ],
      "env": {
        "BACKLOG_DOMAIN": "your-domain.backlog.com",
        "BACKLOG_API_KEY": "your-api-key"
      }
    }
  }
}

Description

Backlog is Nulab's project and issue tracker, and this is the company's own MCP server for it. It exposes the Backlog API to agents such as Claude Desktop, Cline and Cursor, covering the day-to-day objects a project actually runs on: projects, issues and their comments, versions and milestones, wiki pages, and the Git repositories and pull requests attached to a project. Notification tools round it out.

Built for responses that fit in context

Two design choices matter more than the tool list. Field selection lets a caller ask for only the fields it needs, rather than paying for a full issue object when it wanted a title and a status. Token limiting caps large responses, so a query that happens to match several hundred issues degrades instead of blowing the context window. Both are the sort of thing that only becomes obvious after running an agent against a real tracker with years of history in it.

Tool descriptions are also customisable — --export-descriptions writes them out so a team can retune the wording that drives when an agent reaches for each tool.

Getting it running

Docker is the documented default (ghcr.io/nulab/backlog-mcp-server, pulled fresh each run), with npx backlog-mcp-server as the lighter alternative and a manual Node.js setup for contributors. Either way it needs two environment variables: BACKLOG_DOMAIN (your your-domain.backlog.com host) and BACKLOG_API_KEY, issued from your Backlog account.

MIT licensed, maintained by Nulab in the open, and documented in both English and Japanese.

Covered in the Weekly

Related MCP Servers

MCP: Teamwork.com MCP

by Teamwork.com

Teamwork.com's official MCP server for projects, tasks, time, Desk tickets, Spaces and Chat — with profile endpoints that trim the tool list to one role and an optional read-only mode.

Productivity

MCP: ClickUp

by ClickUp

Featured

ClickUp's official remote MCP server — around 40 tools across tasks, search, Docs, Chat and time tracking, over OAuth 2.1 with PKCE.

Productivity

MCP: Shortcut MCP Server

by Shortcut

Shortcut's hosted MCP server: find, create and update stories, epics, iterations and docs from Claude, Cursor or VS Code, authorised by OAuth rather than an API token.

Productivity
Browse all MCP servers →