For teams building with LangGraph

Deploy stateful LangGraph agents to your cloud without Kubernetes

Use Docker Compose and the Defang CLI to run your LangGraph graphs, APIs, and workers on real infrastructure in AWS or GCP with durable storage and TLS handled for you.

Docker Compose-native
Built for LangGraph stacks
BYOC on AWS or GCP
CLI

Deploy the stack from your repo in one command.

defang compose up --provider=aws
Defang launches you to any cloud
Diagram showing a LangGraph node graph deployed as services in the cloud with Defang.
TLS

HTTPS, secrets, env wiring

Included by default when you deploy

Stack fit

Infrastructure that matches LangGraph's durability

LangGraph is designed for long‑running, stateful agents. Defang gives you the infrastructure foundation they need: predictable deployments, repeatable environments, and storage that survives restarts and rollouts.

Durable storage + vector toolsPromote the same graph across envs
01

Model your whole stack in Compose

Capture your LangGraph API, workers, Postgres, and vector store in a single docker-compose.yml. Defang treats it as the blueprint for every environment, from a local dev setup to production in AWS or GCP.

02

Use Bedrock and Vertex AI directly

When your LangGraph agents run inside your own AWS or GCP accounts with Defang, they can call managed LLM endpoints such as Amazon Bedrock or Vertex AI through your existing cloud configuration, with no extra network hops or bespoke gateways.

03

BYOC for sensitive workloads

Your LangGraph agents often touch internal data and APIs. With Defang BYOC, all compute, storage, and logs live inside your own AWS or GCP accounts so you can meet security and compliance requirements.

Path to production

How it works

Defang does for your LangGraph deployment what LangGraph does for your agent orchestration: it turns complexity into a graph of well‑defined steps driven by a single source of truth.

1

Define your LangGraph services

Package your LangGraph app (API server, worker, or both) into containers, then add supporting services like Postgres and a vector store. Use a Compose file to declare how they connect.

2

Deploy with the Defang CLI

From the same repo you use for development, run:

defang compose up --provider=aws

Defang builds the images, provisions infrastructure, configures HTTPS, and exposes your LangGraph entrypoints at stable URLs.

3

Iterate with confidence

Evolve your graphs and tools while keeping the same deployment workflow. Spinning up staging, canary, or per‑tenant environments is as simple as reusing the Compose file with different configuration.

Start from an Agentic LangGraph template

The Defang LangGraph sample showcases a graph‑based agent that can call tools like Tavily Search. It ships with a ready‑made Compose file and a one‑click "Deploy with Defang" button so you can see the full flow end‑to‑end before wiring in your own tools and graphs.

Open the LangGraph sample on GitHubUse it as a blueprint for your own long‑running LangGraph agents.

FAQ