memorybase 0.1.0

A self-hosted project and task management server for AI coding agents

AgentDock

The project operating system for AI agents.

AgentDock gives your AI coding assistants a shared memory and task system. Agents read your project blueprint, update task statuses, record decisions, and coordinate with each other — all through a standardised MCP interface.

---

Key Features

• 17 MCP tools — projects, blueprints, tasks, vault — works with Claude Code, Cursor, and any MCP-compatible client
• Real-time Kanban board — drag-and-drop tasks, live updates via Server-Sent Events when agents make changes
• Blueprint editor — structured project definition with in-place editing (tech stack, folder layout, API specs, constraints)
• Vault explorer — searchable, filterable log of decisions, error patterns, and architectural notes
• Zero-config defaults — SQLite out of the box; PostgreSQL for production
• Single Docker image — frontend and backend ship together

---

Quick Start (pip)

pip install memorybase
memorybase --port 8000

Open http://localhost:8000 — that's it.

---

Quick Start (Docker)

# 1. Run AgentDock
docker run -p 8000:8000 \
  -v $(pwd)/data:/app/data \
  -e AGENTDOCK_API_KEY=my-secret-key \
  ghcr.io/sabbiramin113008/agentdoc:latest

# 2. Open in browser
open http://localhost:8000

That's it. SQLite data is persisted in ./data.

---

Quick Start (Local Dev)

# Clone
git clone https://github.com/sabbiramin113008/agentdoc.git
cd agentdoc

# Start backend (creates venv, installs deps, starts with --reload)
./start-backend.sh

# In a second terminal — start frontend dev server
./start-frontend.sh

• Backend: http://localhost:8000
• Frontend: http://localhost:5273
• API Docs: http://localhost:8000/docs

---

MCP Configuration

Connect any MCP-compatible AI agent to AgentDock:

Claude Code / Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agentdock": {
      "type": "sse",
      "url": "http://localhost:8000/mcp/sse"
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "agentdock": {
      "url": "http://localhost:8000/mcp/sse"
    }
  }
}

Your agent can now call tools like list_projects(), get_blueprint(), create_task(), update_task_status(), add_vault_entry(), and more. See MCP Tools Reference.

---

Screenshots

Kanban board screenshot — coming soon

---

Tech Stack

Layer	Technology
Backend	Python 3.11, FastAPI, SQLModel, SQLite / PostgreSQL
Real-time	Server-Sent Events (sse-starlette)
Agent Protocol	MCP (FastMCP)
Frontend	React 18, Vite, Tailwind CSS v3, shadcn/ui
State	React Query, Zustand
Routing	React Router v6
Drag & Drop	@dnd-kit/core
Docs	MkDocs Material

---

Project Structure

agentdoc/
├── backend/          # FastAPI app, SQLModel models, MCP tools, routers
├── frontend/         # React + Vite + Tailwind frontend
├── docs/             # MkDocs documentation source
├── tests/            # pytest smoke tests
├── Dockerfile        # Multi-stage build (node → python)
├── docker-compose.yml
├── start-backend.sh  # Dev convenience script
└── start-frontend.sh # Dev convenience script

---

Contributing

See CONTRIBUTING for dev setup, project structure, and how to add new MCP tools.

---

License

MIT © sabbiramin113008