jamot-mcp — AI-to-Human Task Coordination MCP Server

A remote MCP server that lets AI agents assign tasks, check team workload, decompose complex instructions, and hand over full context to human contributors — all through a single SSE endpoint.

MCP Server Name: jamot-mcp
Transport: SSE
Endpoint: https://your-server:3001/sse

What This MCP Server Does

jamot-mcp is a task coordination MCP server that exposes 15 tools for AI agents to:

• Create and assign tasks to human team members

• Check workload before assigning (warns if someone is overloaded)

• Decompose complex instructions into subtasks automatically

• Attach full context (chat summary, goals, documents) to every task

• Remember decisions and preferences across conversations

• Suggest workload redistribution when the team is unbalanced

Related MCP server: WritBase

Quick Start

Run with Docker

docker run -d \
  -e MONGO_URI=mongodb+srv://user:password@cluster.mongodb.net/yourdb \
  -e WORKLOAD_THRESHOLD=5 \
  -p 3001:3001 \
  jamot/jamot-mcp:latest

Add to Your AI Platform

LibreChat (librechat.yaml):

mcpSettings:
  allowedDomains:
    - 'your-server'

mcpServers:
  jamot-mcp:
    type: sse
    url: http://your-server:3001/sse
    timeout: 60000

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "jamot-mcp": {
      "url": "http://your-server:3001/sse"
    }
  }
}

Environment Variables

MCP Tools

Task Management

Task Decomposition

Workload & Analytics

Users

Memory

Recommended Agent Instructions

You are a task coordination agent connected to jamot-mcp.

RULES:
1. At the start of every conversation, call get_memory() to recall context.
2. Before assigning any task, always call check_workload_before_assign first.
3. Always use tools — never answer from general knowledge.
4. After important decisions, call save_memory() to persist them.
5. If someone seems overwhelmed, proactively call suggest_redistribution().

Database Requirements

Requires MongoDB with these collections:

• users — team members (read-only, queried for assignments)

• tasks — created and managed by this MCP server

• agent_memory — auto-created for agent long-term memory

Built With

• FastMCP 3.x — MCP server framework

• Motor — async MongoDB driver

• Python 3.11

License

MIT — built by Jamot