MCP Shrimp Task Manager
An MCP server that turns a request into a step plan with dependencies and keeps context across agent sessions
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- The server runs locally and reads and writes files in the data directory
- It gives the agent planning tools that drive the course of the work
Install
Manual install
git clone https://github.com/cjo4m06/mcp-shrimp-task-manager.git
cd mcp-shrimp-task-manager
npm install
npm run buildNeeds Node.js 18 or newer. After the build the entry file is at dist/index.js.
This is third-party code. Review the repository files before installing.
What it does
The server gives an agent a toolkit for planning and tracking work over the MCP protocol. From a natural language request it analyzes requirements, splits large work into linked subtasks, tracks dependencies and status, and stores results in a data directory so progress survives a new session. It has a research mode for systematic technology exploration and a reflection mode for revisiting a task. Project rules are defined separately and applied across tasks. It also ships a web task viewer and a light built-in GUI.
Who it is for. For developers and leads who run large tasks through a coding agent and want to keep the plan and context.
Good fit when
- A large task needs to be split into subtasks with dependencies
- You want the plan and status to persist across agent sessions
- You need sequential task execution or a research mode for a topic
Not a fit when
- You need a one-off edit without planning and task tracking
- You cannot build the project locally or run Node.js
Example request
Plan the task: add JWT authentication, split it into subtasks and show the dependenciesLimitations
It needs Node.js 18 or newer, and the repository must be cloned and built before use. The DATA_DIR variable is required and must be an absolute path, otherwise tasks have nowhere to live. The web viewer and GUI are enabled separately.
How to disable. Remove the server entry from your client's MCP configuration and, if you wish, delete the built project folder.
MCP
- Transport
- stdio
- Authentication
- not required
| Environment variables | |
|---|---|
| DATA_DIR required | Absolute path to the directory that stores tasks and history |
| TEMPLATES_USE | Language of the prompt templates, for example en or zh |
| ENABLE_GUI | Enables the light web overview interface |
| WEB_PORT | Port for the web interface |
Security check
- The server runs locally and reads and writes files in the data directory
- It gives the agent planning tools that drive the course of the work
README in short
The README describes an MCP server for managing tasks in agent-assisted development: planning, decomposition, dependency tracking and memory across sessions. Installation is by cloning the repository, running npm install and npm run build, then adding the server to the MCP config as node pointing to dist/index.js. The required DATA_DIR variable sets the data directory; there are research and reflection modes and ready commands like plan task and execute task. A web task viewer and a built-in GUI are available separately. MIT licensed.
FAQ
Where are tasks stored?
In the directory from the DATA_DIR variable. The path must be absolute; the plan and history are loaded from there in a new session.
Is the web interface required?
No. The server works without it; the task viewer and GUI can be enabled separately if you want a visual overview.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent