Skip to content

Getting Started

Prerequisites

  • Python 3.11+
  • Node.js 18+
  • Git

Option A — pip (simplest)

pip install memorybase
memorybase --port 8000

Open http://localhost:8000. SQLite database is created in the current directory automatically.

Set a secure API key before exposing to a network:

MEMORYBASE_API_KEY=my-secret memorybase --port 8000

Option B — Docker

# 1. Pull the image
docker pull ghcr.io/sabbiramin113008/memorybase:latest

# 2. Create a data directory for SQLite persistence
mkdir -p memorybase_data

# 3. Run
docker run -p 8000:8000 \
  -v $(pwd)/memorybase_data:/app/data \
  -e MEMORYBASE_API_KEY=change-me \
  ghcr.io/sabbiramin113008/memorybase:latest

Open http://localhost:8000.

Option C — Local Development

1. Clone the repo

git clone https://github.com/sabbiramin113008/memorybase.git
cd memorybase

2. Start the backend

./start-backend.sh

This script:

  • Creates and activates a Python virtual environment at ./venv
  • Installs all dependencies from backend/requirements.txt
  • Starts the FastAPI server with --reload on http://localhost:8000

3. Start the frontend

In a second terminal:

./start-frontend.sh

This script:

  • Installs npm dependencies if node_modules is missing
  • Starts the Vite dev server on http://localhost:5273

The frontend proxies all /api, /events, and /mcp requests to the backend automatically.

First Project Walkthrough

1. Open the UI

Navigate to http://localhost:5273 (dev) or http://localhost:8000 (Docker).

2. Create a project

Click New Project, fill in a name, domain, and description, then click Create Project.

3. Define a blueprint

Click Blueprint in the sidebar. Hover over any section and click Edit to fill in your tech stack, folder structure, and constraints.

4. Add tasks

Click Kanban Board. Use the Add Task button (or the + link at the bottom of any column) to create tasks. Drag cards between columns to update status.

5. Connect an AI agent

Go to Settings and copy the MCP config for your tool:

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

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

Once connected, your agent can call tools like list_projects(), get_blueprint(), update_task_status(), and more.

See MCP Tools for the full reference.

Environment Variables

See Configuration for all available settings.

The most important ones to set before going to production:

MEMORYBASE_API_KEY=your-secret-key-here
DATABASE_URL=postgresql://user:pass@host:5432/memorybase