home / mcp / shrimp mcp server

Shrimp MCP Server

Shrimp Task Manager is a task tool built for AI Agents, emphasizing chain-of-thought, reflection, and style consistency. It converts natural language into structured dev tasks with dependency tracking and iterative refinement, enabling agent-like developer behavior in reasoning AI systems.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "cjo4m06-mcp-shrimp-task-manager": {
      "command": "node",
      "args": [
        "/path/to/mcp-shrimp-task-manager/dist/index.js"
      ],
      "env": {
        "DATA_DIR": "/path/to/shrimp_data",
        "ENABLE_GUI": "false",
        "TEMPLATES_USE": "en"
      }
    }
  }
}

MCP Shrimp Task Manager provides persistent memory, structured workflows, and smart task decomposition to help AI agents plan, execute, and verify software development work without losing context. It enables you to break down complex projects into manageable tasks, track progress across sessions, and maintain coding standards as you progress.

How to use

You interact with Shrimp Task Manager through an MCP client to plan and execute tasks. Start by initializing project standards, then ask the system to plan tasks, and finally execute them either step by step or in continuous mode. The system maintains context across sessions, so you can return to tasks later without redoing prior work.

How to install

Prerequisites: Node.js 18+ and npm or yarn. You also need an MCP-compatible AI client such as Claude Code.

Install Claude Code globally and start it, then install Shrimp Task Manager and build it locally.

Install Claude Code (example commands):

# Windows 11 with WSL2
wsl --install
wsl -d Ubuntu
npm install -g @anthropic-ai/claude-code
claude

Install Shrimp Task Manager by cloning the project, installing dependencies, and building the project.

git clone https://github.com/cjo4m06/mcp-shrimp-task-manager.git
cd mcp-shrimp-task-manager
npm install
npm run build

Create a configuration file for the MCP client to run Shrimp Task Manager as a local server. This example shows how to point the MCP client to the built index and provide necessary environment settings.

{
  "mcpServers": {
    "shrimp_task_manager": {
      "command": "node",
      "args": ["/path/to/mcp-shrimp-task-manager/dist/index.js"],
      "env": {
        "DATA_DIR": "/path/to/your/shrimp_data",
        "TEMPLATES_USE": "en",
        "ENABLE_GUI": "false"
      }
    }
  }
}

Start the MCP client with your MCP configuration

Once you have the configuration file, start the MCP client with the custom MCP setup to connect Shrimp Task Manager.

claude --dangerously-skip-permissions --mcp-config .mcp.json

Configuration notes

Environment variables shown for the MCP server are used to customize data storage, templates language, and GUI behavior. You can adjust these values to fit your project structure and preferences.

Start Using

1) Initialize your project standards with init project rules. 2) Plan a task with plan task: <description>. 3) Execute tasks with execute task or switch to continuous mode to run all subtasks automatically. 4) Monitor progress and adjust as needed using the Task Viewer or the Web GUI if enabled.

Environment and commands

Create a local environment file to customize behavior. Example variables include the data directory, GUI enablement, and language templates.

# .env
DATA_DIR=/path/to/data/storage
ENABLE_GUI=true
WEB_PORT=3000
PROMPT_LANGUAGE=en

Tips and quick setup

If you want to view tasks in a modern web interface, enable the GUI by setting ENABLE_GUI=true in the environment. You can access the Task Viewer on the specified port after starting the server.

Troubleshooting

If the MCP client cannot connect to Shrimp Task Manager, verify that the path to the built index.js is correct, the environment variables are set, and that the MCP client is launched with the proper configuration file. Check that Node.js is version 18 or higher and that dependencies are installed successfully.

Notes

Shrimp Task Manager provides persistent memory, structured workflows, and task memory to keep you on track across sessions. It decomposes complex tasks into atomic subtasks and preserves context even when token limits are reached.

Available tools

Task Viewer

Modern React interface for visual task management with drag-and-drop, real-time search, and multi-profile support.

Web GUI

Optional lightweight web interface for quick task overview; enable via ENABLE_GUI in the environment.

Research Mode

Systematic exploration of technologies and solutions to inform task planning.

Agent System

Assign specialized AI agents to specific tasks to distribute work effectively.

Project Rules

Define and enforce coding standards and project-wide guidelines.

Task Memory

Automatic backup and restoration of task history for persistent context.

Plan Task

Create structured task plans from high-level goals.

Execute Task

Carry out tasks and report progress back to the planner.