MCP Kanban Memory

An MCP tool set providing internal task management state for complex multi-session workflows with AI agents. This is similar to other memory MCP tools, with the additional structure, rules and visibility of a kanban-based task management system.

The main idea is to direct the AI agent to document and save its work in a kanban board, both in the planning session and in execution sessions.

Highlights

• Column capacity / work-in-progress limits
• Embedded DB (SQLite)
• Web UI for observing the progress of the workflow, and for modifying tasks manually
• Predefined prompts for starting and resuming a workflow

Usage

Use the MCP prompts to start a project or to make progress on a project. Alternatively, ask the LLM assistant to record its plan by creating a kanban board. To make progress in a follow-up session, ask the assistant to locate a specific kanban board and resume work on it.

Installation

Clone this repo, then run:

npm ci --prefix shared/db
npm ci --prefix mcp-server
npm run build --prefix shared/db
npm run build --prefix mcp-server

Then add the MCP server configuration to the MCP client (e.g. Claude Desktop):

"mcpServers": {
    "kanban-mcp": {
        "command": "node",
        "args": [
            "/path/to/repo/mcp-server/dist/server.js"
        ],
        "env": {
            "MCP_KANBAN_DB_FOLDER_PATH": "/path/to/db"
        }
    }
}

Replace /path/to/repo with the location of the cloned repo.

Replace /path/to/db with a folder that will contain the DB files. This can be any folder with read/write access. It will be created if it doesn't exist.

Installation with Docker

Build the Docker image:

docker build -t mcp/mcp-kanban .

Then add the MCP server configuration to the MCP client (e.g. Claude Desktop):

"mcpServers": {
    "kanban-mcp": {
        "command": "docker",
        "args": [
            "run",
            "--rm",
            "-i",
            "-v",
            "/path/to/db:/mcp",
            "mcp/mcp-kanban"
        ]
    }
}

Replace /path/to/db with a folder that will contain the DB files. This can be any folder with read/write access. It will be created if it doesn't exist.

Using the web UI

Build the web-ui and web-server:

npm ci --prefix shared/db
npm run build --prefix shared/db

npm ci --prefix web-ui
npm ci --prefix web-server
npm run build --prefix web-ui
npm run build --prefix web-server

Run the web-server:

MCP_KANBAN_DB_FOLDER_PATH=/path/to/db npm run start --prefix web-server

Open the browser at http://localhost:8221

API

Tools

• create-kanban-board

  • Create a new kanban board to plan and keep track of your tasks.
  • Input:
    • name (string): The name of the board
    • projectGoal (string): The goal of the project.

• add-task-to-board

  • Add a new task to the landing column (to-do) of a kanban board.
  • Input:
    • boardId (string): The ID of the board to add the task to
    • title (string): The title of the task
    • content (string): The content of the task in markdown format

• move-task

  • Move a task from one column to another, respecting WIP limits.
  • Input:
    • taskId (string): The ID of the task to move
    • targetColumnId (string): The ID of the column to move the task to
    • reason (string, optional): The reason for moving the task

• delete-task

  • Delete a task.
  • Input:
    • taskId (string): The ID of the task to delete

• get-board-info

  • Get the full info of a kanban board, including columns and tasks.
  • Input:
    • boardId (string): The ID of the board to get info for
  • Return detailed information about the board, including columns and tasks.

• get-task-info

  • Get the full info of a task, including its content.
  • Input:
    • taskId (string): The ID of the task to get info for
  • Return detailed information about the task, including its content.

• list-boards

  • List all kanban boards in the database.
  • Input: None
  • Return a list of all boards in the database with their names, creation times, and goals.

Prompts

• create-kanban-based-project

  • Create a kanban board for a project, ask questions to divide the project into tasks, and add them to the board.
  • Input:
    • description (string): The description of the project

• make-progress-on-a-project

  • Make progress on an existing project by locating its kanban board, selecting the next task, and working on it.
  • Input:
    • description (string): The description of the project