Defang

A one-sentence summary

Defang is a cloud deployment platform that uses Docker Compose as its foundation; it enables the deployment of applications to users’ own AWS, GCP, Azure, or DigitalOcean accounts through CLI, MCP, and agent technologies, and it makes use of AI to assist in generating configurations and diagnosing failures.

Tool Introduction

Defang is operated by Defang Software Labs Inc., and its goal is to enable developers to carry out multi-cloud deployments using similar Compose definitions, thereby avoiding the need to manually configure the numerous infrastructure settings for each cloud platform. It handles the creation of containers, as well as the configuration of load balancing, networking, logging, roles, and certain managed services.

The focus of this product is not on generating business code, but rather on delivering existing applications to the cloud in a reliable manner. AI capabilities are used to create project or Compose structures, to diagnose deployment issues, and to work together with coding agents to carry out cost estimates, deployments, status checks, and log reviews.

Product composition

Core deployment capabilities

From Compose to cloud resources

Defang reads the compose.yaml file located in the project’s root directory, and converts the information contained therein – such as services, images, build contexts, ports, resource allocations, and replicas – into cloud resources. For existing Docker Compose projects, it is usually sufficient to adjust only those fields that are not supported, without the need to redescribe the entire application.

Bring Your Own Cloud Account BYOC

The BYOC model involves deploying runtime resources in the customer’s own cloud account, which allows them to make use of their existing budget, permissions, network infrastructure, and compliance requirements. Defang is responsible for control and orchestration, but the customer remains accountable for the costs related to computing power, storage, databases, traffic, and models provided by the cloud service provider.

Construction and rolling updates

During deployment, it is possible to use existing container images, or upload a build context to generate an image in the cloud. Pro and Enterprise support Rolling Deployments, while the Affordable mode of Starter may remove the old service first before creating a replacement service.

Managed databases and caching

The official website’s package listings include services such as hosting for Postgres and Redis; Defang can create the corresponding cloud services based on Compose configurations. Database backup, retention, upgrading, networking, and costs still need to be verified according to the target cloud platform.

GPU and automatic scaling

All plans – Starter, Pro, and Enterprise – list GPU support and automatic scaling capabilities, but the actual available instances, regions, quotas, and costs depend on the target cloud account. It is necessary to check the capacity and cost estimates before deploying large models or GPU services.

Supported cloud platforms

The Preview labels in the public documentation may change depending on the progress made by the provider. Before proceeding with the actual purchase, it is necessary to verify once again that the target region, the services required, the domain names, the databases, and the high availability features are indeed supported.

Three deployment modes

Affordable, Balanced, and High Availability are the built-in deployment options provided by Defang; they allow trade-offs between cost, availability, and redundancy. This document recommends using Stack to record the provider, region, and mode, although the old mode parameter is still usable.

AI generation feature

Defang Generate converts natural language requests into a project outline that can be further modified, and it generates the necessary application files, Dockerfile, and Compose configurations. The Starter version allows for a limited number of generations per month, while the Pro and Enterprise versions offer unlimited Compose generation.

• It can describe the language, frameworks, endpoints, number of services, and dependent components.
• It is possible to create combinations of simple Web services, APIs, queues, Redis, or databases.
• The generated results will be saved in the local project, allowing users to make further modifications in the editor.
• For complex requirements, ports, health checks, environment variables, and data persistence must be clearly defined.
• AI outputs must first be tested locally, with checks carried out on dependencies, security, and cloud resource configurations.
• The limit on the number of generations applies only to the Defang AI functions and does not mean that cloud deployment resources are free.

AI deployment and debugging

When the service resulting from the compose up process fails to reach a healthy state, AI debugging is available for plans at Pro level and above. The intelligence system reads the deployment logs and project files, analyzes reasons for build failures, health check issues, or runtime errors, and provides repair suggestions in the terminal.

• AI debugging is triggered after a failed deployment and cannot be run manually at any time.
• The user’s permission is requested before starting, and sensitive files should not be sent for analysis by default.
• It will suggest modifications, but it will not automatically change the project files.
• The suggestions may be incomplete; before using them, it is necessary to understand the differences and conduct further testing.
• If the logs contain keys, personal data, or customer content, they should be cleaned up first and access to them restricted.
• This feature requires the Pro plan at $49 per month or the Enterprise package.

MCP and agent skills

The official repository includes an MCP Server, and it offers skills or plugins tailored for intelligent agent environments such as Codex and Claude Code. Developers can carry out tasks like cost estimation, deployment, status checking, and log analysis within conversations, without having to memorize all the commands manually.

Complete deployment process

1. Prepare applications that can run locally, along with the Dockerfile and compose.yaml files.
2. Install Defang CLI, log in to your account, and select the correct workspace.
3. Configure credentials, permissions, region, and required service quotas for the target cloud account.
4. Create a Stack to record the cloud provider, region, and mode: Affordable, Balanced, or HA.
5. Use configuration commands to save the keys; do not write the actual secrets directly into Compose or the version control system.
6. First, run the cost estimation to check the calculations, database usage, load balancing, and the main monthly expenses.
7. Execute the deployment and wait for building, resource creation, and health checks to complete.
8. Check the service status, endpoints, and logs, and verify the functions using actual requests.
9. Configure domain names, certificates, monitoring, alerts, backup, and rollback strategies.
10. Continuously monitor Defang subscriptions and cloud provider bills, and delete unused stacks in a timely manner.

Beginner’s guide

1. First, select an official example or a Compose project that contains only a single service.
2. Start it locally using Docker Compose, and verify that the ports and health checks are working properly.
3. After logging in to Defang, use the Playground to get familiar with the commands; do not connect to a production account right from the start.
4. Generate or check compose.yaml, and remove fields that are not yet supported by Defang.
5. Deploy to Playground and view the status, logs, and public endpoints.
6. Switch to BYOC when preparing for production, and select the target cloud along with the most cost-effective deployment configuration.
7. Verify the resources using cost estimates and the cloud console before proceeding with the actual deployment.
8. After successful verification, the same steps are incorporated into CI, and approval processes are established for destructive changes.

Compose support and limitations

Cost estimation

Defang can estimate the main monthly cloud costs based on the services, resource reservations, replicas, providers, and deployment patterns specified in Compose. The current official table shows that AWS and GCP support such estimation, while Azure and DigitalOcean are not yet listed as supporting it.

Estimates do not cover all pay-as-you-go costs, such as data transmission, actual storage growth, and certain third-party services. Before going live, it is necessary to use the cloud provider’s pricing tools, budget alerts, and assumptions based on actual traffic levels.

Which users are it suitable for

• Application developers: Want to quickly move to the cloud using Compose, without having to write a large amount of infrastructure code.
• Small technical teams: need to run services within their own cloud accounts and manage infrastructure costs.
• Startups: Begin with Affordable, and switch to Balanced once traffic increases.
• DevOps team: Aiming to standardize the entry points for multi-cloud deployment, as well as logging and automation processes.
• AI development team: It is necessary to deploy GPUs, host LLMs, as well as provide API and vector or caching services.
• Software suppliers: Need to deliver similar applications to multiple customer cloud accounts.
• Developers using Codex or Claude Code: those who wish to carry out estimation and deployment from within intelligent session environments.

Typical use cases

• Deploy Node.js, Python, Go, or other containerized web services to AWS or GCP.
• Connect multi-service Compose projects with hosted Postgres, Redis, and load balancing.
• Reuse nearly identical application definitions across different cloud platforms to reduce platform-specific configurations.
• Deploy AI inference applications on cloud provider-hosted models or GPU instances.
• Continuously build and deploy containerized applications through CI.
• Estimate costs in coding agents, deploy projects, and check failure logs.
• Manage similar software deployments for multiple customer cloud accounts.

Product advantages

• The Compose specifications, familiar to developers, serve as a bridge between applications and infrastructure.
• It supports the BYOC approach for AWS, GCP, Azure, and DigitalOcean.
• Starter is free and includes a cloud account, making it suitable for testing the deployment process first.
• AI-generated diagnostics help reduce the costs associated with initial setup and troubleshooting.
• MCP, Codex skills, and the Claude Code plugin enable agents to directly participate in deployment workflows.
• Three deployment options enable teams to choose between cost and high availability.
• The CLI source code is licensed under the MIT license, which facilitates auditing, reporting of issues, and automated integration.
• All plans allow an unlimited number of team members, with subscriptions billed on a per Workspace basis.

Usage restrictions and precautions

• The Defang subscription fee does not include the costs associated with cloud service resources, so the actual total cost can be much higher than the monthly fee.
• The Starter version comes only in the Affordable mode, and there may be a brief interruption during updates.
• Compose is not fully compatible; top-level volumes and secrets are not supported at the moment.
• depends_on does not guarantee that the hosting service is ready; the application must implement retry logic.
• Playground is intended for testing purposes; it does not support custom domain names and should not be used to store production data.
• The documentation for Azure, GCP, and DigitalOcean still shows a preview indicator; it needs to be reviewed before being used in production.
• AI suggestions may be incorrect; infrastructure or security settings should not be modified directly without review.
• For domain names, only one configuration can be set per service at the moment, and some providers do not support BYOD.
• The cost estimate does not cover all expenses related to data transmission, storage, and pay-as-you-go services.
• Improper configuration of cloud account permissions can lead to exposed resources, uncontrolled costs, or deployment failures.

Prices and packages

As of August 22, 2026, Defang charges based on the number of workspaces, and all plans include an unlimited number of team members. The Starter plan is free, the Pro plan costs $49 per month, and the Enterprise plan costs $499 per month; additional cloud accounts and infrastructure services provided by cloud providers require separate billing.

All three packages include GPU, automatic scaling, and hosted Postgres and Redis. The Pro version adds rolling deployment, while the Enterprise version includes Fleet Management, WAF, deployment across multiple customer cloud accounts, and high-availability mode; the final benefits depend on the settlement terms and contract agreements.

Supported platforms and installation methods

Managed LLM deployment

Defang can map LLM services that have extension tags defined to the hosted model platforms in the target cloud. The current document lists AWS Bedrock, Google Vertex AI, and Microsoft Foundry as supported options; DigitalOcean GenAI is not yet supported.

• Playground can only use the default hosting model selected by Defang.
• BYOC allows the use of models that are authorized in the target cloud account.
• Google Vertex AI requires that access to specific models be enabled in advance.
• OpenAI-compatible applications can adapt to cloud-native models through Access Gateway.
• The model token fees, throughput limits, as well as regional and content policies are determined by the cloud platform.
• Before production, the model version should be fixed, and the output, latency, and fallback options should be tested.

Data security and privacy

Defang’s BYOC goal is to have the runtime environment and key resources located in the customer’s cloud account; according to the official documentation, the code and secrets related to AWS and GCP processes remain within the customer’s infrastructure. The processes related to building applications, managing control interfaces, performing AI-related debugging, and handling data flows specific to other providers still need to be verified in accordance with the architecture and DPA guidelines.

• Create cloud roles with the minimum required permissions to avoid using long-term administrator credentials.
• Write the key into Defang configuration management; do not submit it to Compose or Git.
• Before using AI for debugging, check whether the logs and project files contain customer or personal data.
• The privacy policy states that personal information may be processed in Canada and the United States.
• For corporate purchases, a data processing appendix can be signed, and the subcontractors as well as the deletion processes can be verified.
• The CLI includes analysis capabilities by default; Analytics can be disabled using the official environment variables.
• The Enterprise package lists the compliance capabilities for WAF and SOC 2 Type II.
• Configure auditing and alerts for cloud billing, permission changes, deployment, and log access.

API, CLI, and open-source status

Defang’s official repository contains the CLI, MCP Server, plugins, skills, examples, and related tools, and it is licensed under the MIT license. Developers can view and modify this client code, but the hosting portal, control panel, subscription services, and all cloud-based components are not made available under an open-source license.

Basic information

Recommendation score

Its rating is 4.4 out of 5 points. Defang is very useful for development teams that are already using Docker Compose and wish to deploy their workloads to their own cloud accounts; AI-based debugging and agent integration also lower the barriers to multi-cloud deployment.

The main limitations are that Compose’s support is not complete, costs for cloud resources still apply even outside of subscription plans, and some providers offer only preview versions of their services. Before using it in production, it is necessary to verify the target cloud platform, data services, domain names, as well as the requirements related to high availability.

Frequently Asked Questions

What is Defang?

Defang is a cloud application deployment platform that uses Docker Compose as its foundation; it allows for the deployment of containerized applications to one’s own cloud account via CLI or AI agents, while also enabling the management of status, logs, and updates.

Is Defang free?

There is a free Starter package that includes one cloud account, the Affordable plan, and limited AI generation capabilities. The costs associated with cloud service resources are not included in the free plan.

How much is Pro?

Pro currently costs $49 per month and includes one cloud account, Balanced mode, rolling deployment, unlimited Compose generation, AI debugging, and 1 day of support response. Additional cloud accounts cost $29 per month each.

How much is Enterprise?

Enterprise currently costs $499 per month and includes three cloud accounts, high-availability mode, WAF, Fleet Management, and dedicated support. Additional cloud accounts cost $49 per month each.

Are the costs for cloud servers included in the subscription?

It is not included; BYOC resources are billed directly by AWS, GCP, Azure, or DigitalOcean. It is necessary to run estimates prior to deployment and set up cloud budget alerts.

Which cloud platforms are supported?

The current document lists AWS, GCP, Azure, and DigitalOcean. Some platform pages still indicate a \"Public Preview\" status; it is necessary to verify the target service and region before using them in production.

Is it necessary to know Docker Compose?

It is advisable to have basic knowledge of Compose and containers. While AI can generate an initial project, users still need to understand ports, health checks, environment variables, resources, and data persistence.

Does AI debugging automatically modify the code?

No, it reads the logs and project files after a deployment failure and provides suggestions; it is up to the user to decide whether to apply them. This feature is available only in Pro or Enterprise versions.

Does Defang support Codex?

Support is available; the official repository provides methods for installing the Codex plugin and Agent Skills, which enable estimation, deployment, and status checking. For actual use, it is still necessary to install Defang CLI and log in.

Is Defang open source?

The main CLI repository is licensed under the MIT license and contains MCP and plugin code. The Defang-hosted backend, Portal, and commercial platform cannot therefore be considered fully open source.

Can the database be run?

Services such as hosted Postgres and Redis that are supported by Defang can be used, but top-level Compose volumes are not supported at the moment. Production databases should have backup, encryption, and upgrade strategies in place.

Is it suitable for personal projects?

It is suitable for individual developers who are familiar with containers and wish to learn about cloud deployment; the Starter plan is sufficient for testing out the process. Paid plans should be considered when continuous online availability, high reliability, or frequent AI debugging are required.

Summary

Defang combines Compose, BYOC, cloud resource orchestration, and AI-driven connectivity to provide developers with a shorter path from local containers to multi-cloud production environments. Its free Starter version, MIT CLI, and agent capabilities are suitable for initial testing on a small scale.

When using it in a formal setting, it is necessary to take into account factors such as Defang subscriptions, cloud resources, data services, traffic volume, and operational costs; the deployment model should be chosen based on the tolerance level for downtime. While AI can speed up the process of generation and debugging, it cannot replace security and infrastructure assessments.