Quick Start Tutorial

Welcome to Synapse! This tutorial will guide you through installing Synapse and using it to "prime" an AI agent with your project's context.

Prerequisites

Node.js v22.6.0 or higher.
An MCP-compatible client (e.g., Claude Desktop, Cursor, or Gemini CLI).

1. Installation

Install the Synapse CLI globally:

npm install -g synapse-cortex

2. Workspace Initialization

Navigate to a project you are currently working on and run the setup sequence:

synapse setup
synapse doctor

setup: Initializes your local context database.
doctor: Confirms your environment is ready.

3. Connecting to your AI

Add Synapse to your AI client. For Claude Desktop, add this to your configuration file:

{
  "mcpServers": {
    "synapse": {
      "command": "synapse",
      "env": {
        "MCP_MODE": "stdio",
        "SYNAPSE_MEMORY_ENABLED": "true"
      }
    }
  }
}

4. Your First "Prime"

Start a new chat with your agent and ask:

"Prime yourself with this project and give me a high-level summary of the architecture."

What happens: The agent will automatically call synapse_agent_prime. This tool reads your project structure, existing memories, and recent changes to give the agent an immediate "mental map" of your work.

5. Capturing a Lesson

Tell the agent something important about your workflow:

"Remember that for this project, we always use functional components for React and avoid class components."

What happens: The agent uses synapse_memory_manage to store this rule. In your next session, Synapse will automatically remind the agent of this preference.

Next Steps: