Quickstart

Fresh

Source: Letta Code Quickstart

Setup Methods

Desktop App (Recommended)

1. Download and install Letta Code from the official site
2. Available for macOS, Windows, and Linux
3. Launch the app and sign in with a Letta account
4. Connect model providers via the bottom-left menu to add external API keys

CLI Installation

Prerequisites

• Node.js v18 or later

Install

npm install -g @letta-ai/letta-code

Authenticate

letta
# Opens browser for OAuth login
# Credentials stored locally

Start Using

# Navigate to your project
cd my-project

# Start interactive mode
letta

Cloud Deployment

For remote machines, install via SSH then start the server:

letta server

This generates an OAuth URL for browser approval. Once approved, connect from:

• chat.letta.com
• The Letta Code desktop app

First Steps After Setup

Initialize Memory

/init

The /init command performs guided initialization using context constitution principles for durable identity, preferences, and project structure. The agent can analyze prior Claude Code and OpenAI Codex sessions to understand your working patterns.

Essential Commands

Command	Description
letta	Start interactive mode
letta -p "query"	Run headless one-shot query
/init	Initialize agent memory
/memory	Manage memory blocks
/search	Find past messages
/remember	Direct agent to store information
/doctor	Audit and refine memory structure
shift-tab	Toggle between modes

Training Your Agent

The more you use your agent, the more valuable it becomes:

1. Ask the agent about existing knowledge to test understanding
2. Create user profiles to personalize interactions
3. Use /remember to correct mistakes and encourage memory updates
4. Agents can search prior conversation data for context

Best Practice

Always use /remember when the agent makes a mistake or misunderstands something. This explicitly directs the agent to store the correction in memory.

Verification Checklist

• [ ] Letta Code installed (Desktop, CLI, or Cloud)
• [ ] Authentication completed
• [ ] Model provider connected
• [ ] /init run successfully
• [ ] Agent responds to basic queries
• [ ] Memory persists between sessions

Next Steps

• Memory System - Deep dive into MemFS and memory management
• Skills - Extend agent capabilities
• Subagents - Delegate specialized tasks
• Headless Mode - Non-interactive automation