Skip to main content
Glama
Shrimp Task Manager

Shrimp Task Manager

🇺🇸 English | 🇩🇪 Deutsch | 🇪🇸 Español | 🇫🇷 Français | 🇮🇹 Italiano | 🇮🇳 हिन्दी | 🇰🇷 한국어 | 🇧🇷 Português | 🇷🇺 Русский | 🇨🇳 中文

MCP Shrimp Task Manager

🦐 Intelligent task management for AI-powered development - Break down complex projects into manageable tasks, maintain context across sessions, and accelerate your development workflow.

Shrimp Task Manager Demo

Watch Demo Video • Quick Start • Documentation

🚀 Quick Start

Prerequisites

  • Node.js 18+

  • npm or yarn

  • MCP-compatible AI client (Claude Code, etc.)

Installation

Installing Claude Code

Windows 11 (with WSL2):

# First, ensure WSL2 is installed (in PowerShell as Administrator)
wsl --install

# Enter Ubuntu/WSL environment
wsl -d Ubuntu

# Install Claude Code globally
npm install -g @anthropic-ai/claude-code

# Start Claude Code
claude

macOS/Linux:

# Install Claude Code globally
npm install -g @anthropic-ai/claude-code

# Start Claude Code
claude

Installing Shrimp Task Manager

# Clone the repository
git clone https://github.com/cjo4m06/mcp-shrimp-task-manager.git
cd mcp-shrimp-task-manager

# Install dependencies
npm install

# Build the project
npm run build

Configure Claude Code

Create a .mcp.json file in your project directory:

{
  "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"
      }
    }
  }
}

Example configuration:

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

Then start Claude Code with your custom MCP configuration:

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

Cline (VS Code Extension): A VS Code extension for AI-assisted coding. Add to VS Code settings.json under cline.mcpServers

Claude Desktop: Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)

Start Using

  1. Initialize your project: "init project rules"

  2. Plan a task: "plan task: implement user authentication"

  3. Execute tasks: "execute task" or "continuous mode"

Related MCP server: Vibe Coder MCP

💡 What is Shrimp?

Shrimp Task Manager is an MCP (Model Context Protocol) server that transforms how AI agents approach software development. Instead of losing context or repeating work, Shrimp provides:

  • 🧠 Persistent Memory: Tasks and progress persist across sessions

  • 📋 Structured Workflows: Guided processes for planning, execution, and verification

  • 🔄 Smart Decomposition: Automatically breaks complex tasks into manageable subtasks

  • 🎯 Context Preservation: Never lose your place, even with token limits

✨ Core Features

Task Management

  • Intelligent Planning: Deep analysis of requirements before implementation

  • Task Decomposition: Break down large projects into atomic, testable units

  • Dependency Tracking: Automatic management of task relationships

  • Progress Monitoring: Real-time status tracking and updates

Advanced Capabilities

  • 🔬 Research Mode: Systematic exploration of technologies and solutions

  • 🤖 Agent System: Assign specialized AI agents to specific tasks (Learn more)

  • 📏 Project Rules: Define and maintain coding standards across your project

  • 💾 Task Memory: Automatic backup and restoration of task history

Web Interfaces

🖥️ Task Viewer

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

Quick Setup:

cd tools/task-viewer
npm install
npm run start:all
# Access at http://localhost:5173

📖 Full Task Viewer Documentation

🌐 Web GUI

Optional lightweight web interface for quick task overview.

Enable in .env: ENABLE_GUI=true

📚 Documentation

🎯 Common Use Cases

Agent: "plan task: add user authentication with JWT"
# Agent analyzes codebase, creates subtasks

Agent: "execute task"
# Implements authentication step by step
Agent: "plan task: fix memory leak in data processing"
# Agent researches issue, creates fix plan

Agent: "continuous mode"
# Executes all fix tasks automatically
Agent: "research: compare React vs Vue for this project"
# Systematic analysis with pros/cons

Agent: "plan task: migrate component to chosen framework"
# Creates migration plan based on research

🛠️ Configuration

Environment Variables

Create a .env file:

# Required
DATA_DIR=/path/to/data/storage

# Optional
ENABLE_GUI=true          # Enable web GUI
WEB_PORT=3000           # Custom web port
PROMPT_LANGUAGE=en      # Prompt language (en, zh, etc.)

Available Commands

Command

Description

init project rules

Initialize project standards

plan task [description]

Create a task plan

execute task [id]

Execute specific task

continuous mode

Execute all tasks sequentially

list tasks

Show all tasks

research [topic]

Enter research mode

reflect task [id]

Review and improve task

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🌟 Credits

Created by cjo4m06 and maintained by the community.


Available Tools

15 tools
analyze_taskC

深入分析任務需求並系統性檢查代碼庫,評估技術可行性與潛在風險,如果需要提供程式碼請使用 pseudocode 格式且盡量精簡只保留核心實現部分

Inputs: summary (required), initialConcept (required), previousAnalysis.

clear_all_tasksA

刪除系統中所有未完成的任務,該指令必須由用戶明確確認才能執行

Inputs: confirm (required).

complete_taskB

正式標記任務為完成狀態,生成詳細的完成報告,並更新關聯任務的依賴狀態

Inputs: taskId (required), summary.

delete_taskA

刪除未完成的任務,但不允許刪除已完成的任務,確保系統記錄的完整性

Inputs: taskId (required).

execute_taskC

按照預定義計劃執行特定任務,確保每個步驟的輸出符合質量標準

Inputs: taskId (required).

get_task_detailA

根據任務ID獲取任務的完整詳細信息,包括未截斷的實現指南和驗證標準等

Inputs: taskId (required).

list_tasksA

生成結構化任務清單,包含完整狀態追蹤、優先級和依賴關係

Inputs: status (required).

plan_taskB

初始化並詳細規劃任務流程,建立明確的目標與成功標準,可選擇參考現有任務進行延續規劃

Inputs: description (required), requirements, existingTasksReference.

process_thoughtA

你可以透過靈活的、可適應和發展的思考過程來分析問題,隨著理解的加深,每個想法都可以建立、質疑或修改先前的見解。你可以質疑想法、假設想法、驗證想法,並且可以建立新的想法。你將重複這個過程,直到你對問題有足夠的理解,並且能夠提出有效的解決方案。如果你覺得思考已經充分可以把 nextThoughtNeeded 設為 false 並且停止思考。

Inputs: stage (required), thought (required), thought_number (required), total_thoughts (required), +4 more.

query_taskB

根據關鍵字或ID搜尋任務,顯示省略版的任務資訊

Inputs: query (required), isId, page, pageSize.

reflect_taskB

批判性審查分析結果,評估方案完整性並識別優化機會,確保解決方案符合最佳實踐,如果需要提供程式碼請使用 pseudocode 格式且盡量精簡只保留核心實現部分

Inputs: summary (required), analysis (required).

split_tasksA

將複雜任務分解為獨立且可追蹤的子任務,建立明確的依賴關係和優先順序。支援四種任務更新模式:追加(append)、覆蓋(overwrite)、選擇性更新(selective)和清除所有任務(clearAllTasks),其中覆蓋模式只會刪除未完成的任務並保留已完成任務,選擇性更新模式可根據任務名稱智能匹配更新現有任務,同時保留其他任務,如果你需要規劃全新的任務請使用清除所有任務模式會清除所有任務並創建備份。請優先使用清除所有任務模式,只有用戶要求變更或修改計畫內容才使用其他模式。

**請參考之前的分析結果提供 pseudocode

**如果任務太多或內容過長,請分批使用「split_tasks」工具,每次只提交一小部分任務

Inputs: tasks (required), updateMode (required), globalAnalysisResult.

update_taskB

更新任務內容,包括名稱、描述和注記,但不允許修改已完成的任務

Inputs: taskId (required), name, notes, description, +4 more.

update_task_filesC

更新任務相關文件列表,用於記錄與任務相關的代碼文件、參考資料等

Inputs: taskId (required), relatedFiles (required).

verify_taskB

全面驗證任務完成度,確保所有需求與技術標準都已滿足,並無遺漏細節

Inputs: taskId (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 15 tools

    Disambiguation4/5

    Most tools have distinct purposes (e.g., delete_task vs complete_task), but analyze_task and reflect_task both involve analysis and could cause confusion. process_thought is a generic cognitive tool that doesn't fit the task management domain. Overall, descriptions help differentiate.

    Naming Consistency5/5

    All tools follow a consistent verb_noun pattern (e.g., list_tasks, update_task, complete_task). Even process_thought adheres to this pattern. No mixing of conventions like camelCase or inconsistent verb styles.

    Tool Count5/5

    With 15 tools, the server is well-scoped for a task manager. Each tool serves a clear role in the task lifecycle, and the count falls within the optimal 3-15 range without feeling overly heavy or thin.

    Completeness4/5

    The tool set covers core task management operations: create (plan_task, split_tasks), read (list_tasks, get_task_detail, query_task), update (update_task, update_task_files), delete (delete_task, clear_all_tasks), and completion (complete_task, verify_task). Missing explicit dependency management or prioritization, but analysis tools (analyze, reflect, execute) add depth. process_thought is an outlier.

    Maintenance

    ActivityInactive
    ResponsivenessMostly unanswered

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      AI-driven task management application that operates via MCP, enabling autonomous creation, organization, and execution of tasks with support for subtasks, priorities, and progress tracking.
      2
      MIT
    • A
      license
      B
      quality
      D
      maintenance
      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.
      15
      199 npm
      21
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      MCP server for task management that enables AI agents to read, create, update tasks, and track work sessions, allowing agents and humans to collaborate on the same task board.
      8 npm
      13
      MIT