CLI Overview

Introduction to the LangGraph CLI for building and deploying stateful, multi-actor agents

LangGraph CLI

The LangGraph CLI is the official command-line interface for LangGraph, providing tools to create, develop, and deploy LangGraph applications locally and to production environments.

What is the LangGraph CLI?

The CLI streamlines your LangGraph development workflow by:

• Creating new projects from templates
• Developing locally with hot reloading and debugging
• Building Docker images for deployment
• Testing your graphs in a production-like environment
• Deploying to cloud platforms or self-hosted infrastructure

Installation

Install the CLI via pip:

pip install langgraph-cli

For development mode with hot reloading and in-memory server support:

pip install "langgraph-cli[inmem]"

The in-memory development server requires Python 3.11 or higher.

Quick Start

1. Create a New Project

Start a new LangGraph project from a template:

langgraph new my-agent --template react-agent

2. Develop Locally

Run your agent with hot reloading:

cd my-agent
langgraph dev

This starts a development server at http://localhost:2024 with:

• Automatic code reloading on file changes
• Interactive API documentation at /docs
• LangGraph Studio integration

3. Test with Docker

Test your agent in a production-like environment:

langgraph up

This launches your agent in Docker with all dependencies properly containerized.

4. Build for Production

Create a Docker image for deployment:

langgraph build -t my-agent:latest

Core Concepts

Configuration File

The CLI uses a langgraph.json configuration file that defines:

• Dependencies: Python packages required by your agent
• Graphs: Entry points to your compiled graph objects
• Environment: Variables and configuration for runtime
• Python Version: Target Python version for deployment

Example langgraph.json:

{
  "dependencies": [
    "langchain-openai",
    "langchain-anthropic",
    "."
  ],
  "graphs": {
    "agent": "./src/agent.py:graph"
  },
  "env": ".env",
  "python_version": "3.11"
}

Development vs Production

The CLI provides two modes for running your agent:

Feature	langgraph dev	langgraph up
Environment	In-memory server	Docker containers
Hot Reload	Yes	Optional with --watch
Setup Time	Fast	Slower (builds image)
Use Case	Development	Testing, Production
Requirements	Python 3.11+	Docker

Command Overview

The CLI provides five main commands:

langgraph new

Create a new project from a template

langgraph dev

Run a development server with hot reloading

langgraph up

Launch the API server in Docker

langgraph build

Build a Docker image for deployment

langgraph dockerfile

Generate a Dockerfile for custom deployments

Typical Workflow

1. Create: Use langgraph new to scaffold a project
2. Develop: Use langgraph dev for rapid iteration with hot reload
3. Test: Use langgraph up to test in Docker before deploying
4. Build: Use langgraph build to create production images
5. Deploy: Push your image to your container registry and deploy

Project Structure

A typical LangGraph project structure:

my-agent/
├── langgraph.json       # CLI configuration
├── .env                 # Environment variables
├── src/
│   ├── agent.py        # Your graph definition
│   └── tools.py        # Agent tools and utilities
├── requirements.txt     # Additional dependencies (optional)
└── README.md

Environment Variables

The CLI respects environment variables defined in:

1. .env files (referenced in langgraph.json)
2. System environment variables
3. Docker Compose environment configurations

Common environment variables:

# Development with LangSmith
LANGSMITH_API_KEY=your-api-key

# Production with license
LANGGRAPH_CLOUD_LICENSE_KEY=your-license-key

# Model providers
OPENAI_API_KEY=your-openai-key
ANTHROPIC_API_KEY=your-anthropic-key

Next Steps

• Learn about all CLI commands and options
• Understand langgraph.json configuration
• Explore deployment options

Getting Help

View help for any command:

langgraph --help
langgraph dev --help
langgraph build --help

Version Information

Check your CLI version:

langgraph --version