Defang Software Labs

Defang lets you build AI agents once and deploy them to any cloud environment with a single command. It handles all the infrastructure, including networking, compute, databases, and LLMs, mapped to native services on AWS, GCP, and beyond. Try it at defang.io.

Connecting GitHub to GCP used to mean: - JSON service account keys - Complex IAM setup - Long GitHub Actions configs Not anymore. With Defang, it’s just a few clicks in the Portal: • Connect your GCP project • Configure trust via OIDC • Push to GitHub That’s it. Your next git push deploys to GCP.

Want to try a sample app on your own cloud? Pick a sample in the Defang Portal. Then connect an AWS or GCP account. Then we template a GitHub repo for you. One flow. No context-switching between docs, GitHub, and the cloud console.

We rewrote our log streaming pipeline. defang logs --follow now streams more reliably with fewer race conditions between subscribe and publish. Same behavior. Cleaner code. More consistent logs when you need them.

See exactly what’s running in your cloud. New deployment detail views include resource tagging and direct links to your cloud console for easier inspection and management.

Deploy with Defang via AWS Marketplace. Subscribe in a few clicks. Roll costs into your existing AWS bill. Keep procurement simple. Everything runs in your AWS account, now it’s even easier to get started.

The Defang Portal now includes a command palette for faster navigation and actions. Press Cmd+K to instantly: • Jump to projects • Open deployments • Trigger common actions • Navigate without leaving your keyboard

Defang is SOC 2 compliant. Security isn’t optional when you’re shipping real infrastructure. See our Trust Center: https://trust.defang.io/

We’ve added support for: defang config set --if-not-set This flag ensures a config value is only set if it hasn’t already been defined. It’s especially useful in CI or automated environments where you want manually applied configuration to take precedence and avoid unintended overwrites.

The Defang CLI login flow has been updated for a smoother login experience. It now uses shorter URLs and logs you into the Portal at the same time. Simpler flow. Fewer steps. Consistent sessions.

See exactly what’s running in your cloud. New deployment detail views include resource tagging and direct links to your cloud console for easier inspection and management.

Deploy to AWS from GitHub Actions using OIDC — no access keys required. Configure trust in the Defang Portal and let GitHub authenticate directly to your AWS account. Cleaner. More secure. Simpler CI.

Kiro generates the app. The Defang Kiro Power turns it into real cloud infrastructure. From vibe-coded Flask app → deployed to AWS ECS in minutes. No Terraform. No IAM guesswork. Build in your IDE. Deploy with one command.

Built something with GPT-5.3 Codex? Use the Defang MCP Server to deploy it directly to your AWS or GCP account. Your AI writes the code. Defang handles the infrastructure. Real cloud. Your account. No boilerplate.

Need dev, staging, production, and per-customer deployments? Stacks let you deploy isolated environments from the same compose file. No config duplication. No infrastructure drift. More info: https://defang.io/blog/post/named-stacks-deploy-multiple-environments-from-one-codebase/

With the Defang MCP setup, Claude Code can deploy directly to AWS or GCP. Run: defang mcp setup --client=claude-code Then prompt: “deploy this with defang” Works seamlessly with newer Claude models like Opus 4.6 for complex, multi-step deployments. Claude builds. Defang provisions.

AI agents run continuously, call LLMs, and manage state. They need more than basic hosting. Your Compose file becomes production infrastructure. AWS and GCP in one command. Learn more: https://defang.io/blog/post/what-s-the-best-infrastructure-for-running-ai-agents-at-scale/

If your GitHub repo is public, Defang lets you deploy to your own AWS or GCP account for free. No trial. No credit card. No approval process. You pay your cloud provider directly. Defang’s fee is zero. Open source deserves production-grade deployments.

Defang lets you use managed Redis across your services without worrying about networking or environment-specific connection strings. Just define it in Compose and Defang provisions and connects it for you on AWS or GCP. More info: https://docs.defang.io/docs/concepts/managed-storage/managed-redis

We're adding new AI features soon. In the meantime check out what you can do with our MCP server! More info here: https://docs.defang.io/docs/concepts/mcp

Defang supports horizontal scaling for cloud-native services and offers auto-scaling on AWS and GCP. Services can automatically scale based on demand, improving performance during traffic spikes and reducing cost when load is low. More info: https://docs.defang.io/docs/concepts/scaling

Missing or broken Dockerfiles are a common deployment failure. Defang avoids this by using Railpack to build images automatically for Node, Python, Go, Java, Ruby, Rust, PHP, Elixir, and static sites. Just bring your app and a Compose file. Read more https://docs.defang.io/docs/concepts/railpack

Defang applies least-privilege access, creating scoped roles for each service. Deployments run in private subnets with controlled ingress through an Application Load Balancer. SSL certificates are automated with ACM. Secrets are stored in Parameter Store and encrypted with KMS.

Defang supports one-off jobs for tasks like migrations, and you can control their order with service dependencies in your Compose setup. Each job waits for the services it relies on, ensuring a clean flow from database to migration to app. More info https://docs.defang.io/docs/concepts/one-off-jobs

Have a Docker compose.yaml? Authenticate your cloud account, then run: defang compose up --provider=aws Or switch out aws for gcp or digitalocean if that's what you use. Defang provisions the resources and deploys your services, all in your own account. Check out docs.defang.io for more info.

Defang keeps networking familiar: public services use load balancers, private services run in isolated subnets, and internal DNS handles service-to-service calls. It’s the Compose networking model applied securely and at scale in your cloud. More info https://docs.defang.io/docs/concepts/networking

Defang follows Twelve-Factor principles: stateless services, env-based config, externalized state, and fast disposable processes that scale cleanly. Learn more: https://docs.defang.io/docs/concepts

Come listen to us live as we talk to the Mastra team about deploying agents to the cloud! #mastra #cloud #agents #ai https://www.youtube.com/watch?v=NyWxvSMALdA

Connect Defang to your IDE using the MCP Server and deploy services through your AI coding agent. Setup: defang mcp setup --client=<your-ide> Supports VS Code, Cursor, Claude, Windsurf, Rovo, and Kiro. Try prompts like “deploy this with defang.” More info: https://docs.defang.io/docs/concepts/mcp

What side project are you deploying this Saturday? #programming

Wiring logs shouldn’t be hard. * `defang compose up` streams build + runtime logs right away * Or jump into `defang tail --name my-service` Deploy first, *then* decide if you ever want to open your cloud provider’s logging console again. 🚢🐙 https://docs.defang.io/docs/concepts/observability

Wiring logs shouldn’t be hard. * `defang compose up` streams build + runtime logs right away * Or jump into `defang tail --name my-service` Deploy first, *then* decide if you ever want to open your cloud provider’s logging console again. 🚢🐙 https://docs.defang.io/docs/concepts/observability

🚀 Major UI update to the Defang Portal. We believe that a smoother, more intuitive interface helps you keep better track of your deployments. You'll see more changes in the near future too. We’d love to hear your thoughts! what stands out? What could still be improved? #cloudops #userexperience

Defang has three deployment modes: Affordable for dev, Balanced for staging, High Availability for production. Each mode adjusts compute, logs, networking, and storage so you only pay for what you need at each stage.

Defang applies least-privilege access, creating scoped roles for each service. Deployments run in private subnets with controlled ingress through an Application Load Balancer. SSL certificates are automated with ACM. Secrets are stored in Parameter Store and encrypted with KMS.

Deploy a LangGraph Agent in one command. Start here: https://github.com/DefangSamples/sample-agentic-langgraph-template Set your Tavily API key: defang config set TAVILY_API_KEY Run locally: docker compose -f compose.dev.yaml up --build Deploy: defang compose up

Connect Defang to your IDE with the MCP Server and deploy straight from your editor. Works with VS Code, Cursor, Claude, Windsurf, and more. Run: defang mcp setup --client=<your-ide> Then try prompts like “deploy this with defang.” Learn more: https://docs.defang.io/docs/concepts/mcp

Deploy a Strands Agent in one command. Start here: https://github.com/DefangSamples/sample-agentic-strands-template 1-click deploy to the Defang Playground or your own cloud (AWS, GCP, DigitalOcean). Run locally: docker compose -f compose.dev.yaml up --build Deploy: defang compose up

Have a Docker compose.yaml? Authenticate your cloud account, then run: defang compose up --provider=aws Or switch out aws for gcp or digitalocean if that's what you use. Defang provisions the resources and deploys your services, all in your own account. Check out docs.defang.io for more info.

Did you know Defang's cost estimation isn’t just a ballpark number. It breaks down resources like load balancers, NAT gateways, managed databases, vCPUs, and storage per service defined in your Compose file. See the real cost of scaling up or down, and optimize your stack for value. #finops #cloud

I'm a big fan of #kubernetes in many ways. I started using it in 2019 and quickly became a fan (kinda): I was able to run my containers across multiple nodes and scale out. But it was overkill for what I was doing, and always required building a bunch of extra tooling to get the behaviour I needed.

Some people see #Kubernetes as *the* way to deploy #containers to the cloud. But it's big, complex, and not as portable as it seems. Kubernetes has it's place, but at #Defang we're building something for the rest of us. For #devs who want simplicity, portability, and a whole lot more. #devops

Did you know that you can tell #Defang that a service requires a GPU? ``` gpu-service: deploy: resources: reservations: devices: - capabilities: ["gpu"] ``` Whether you're deploying to AWS or GCP, we'll make sure to properly configure your service. #GPU

Wish you could compare cloud costs for AWS vs GCP without setting up billing? Just run `defang estimate` and review a cost estimate for your application across providers and deployment modes: affordable, balanced, high availability. #cloud #finops #programming #deploy

Have you tried the #docker model runner? It's a nice way to download LLMs, run them in a #DockerCompose app, directly integrated with the rest of your code. Also, did you know Defang will deploy your app to #AWS or #GCP, replacing the local models with #Bedrock or #Vertex?

How did you bring down prod? Any stories? Share below 👇 (we promise we won't tell anyone) #programming

Sundays are for side projects, right? What are you building today? Share below 👇 #programming

It's Saturday! What side project are you working on? Share below 👇 #programming