AgentOS on AWS

Deploy AgentOS to AWS with ECS Fargate

Deploy AgentOS to AWS with ECS Fargate, RDS PostgreSQL, and an Application Load Balancer. Production-grade infrastructure that runs entirely in your cloud.

Architecture

Cost

Resource	Purpose	Monthly Cost
ECS Fargate	Serverless container hosting	$30-50
RDS PostgreSQL	Managed database with pgvector	$25
Application Load Balancer	HTTP/HTTPS endpoint	$20-25
AWS Secrets Manager	Secure credential storage	< $1
Security Groups	Network access controls	Free
Total		~$75-100

Use the AWS Pricing Calculator for detailed estimates.

---

Prerequisites

1. Install tools

   • Install Docker Desktop
   • Install uv (Python package manager)
   • Install and configure the AWS CLI
2. Create and activate a virtual environment

   uv venv --python 3.12
   source .venv/bin/activate
   
3. Install Agno

   uv pip install -U 'agno[infra]'
   
4. Create your codebase

   ag infra create --template agentos-aws --name my-agentos
   
   cd my-agentos
   

   Or clone directly: git clone https://github.com/agno-agi/agentos-aws-template.git

5. Set your API key

   export OPENAI_API_KEY=sk-***
   

HTTPS is required to connect to os.agno.com. You’ll set this up after deployment.

---

Deploy

Step 1: AWS Setup

Choose your preferred method to create the ECR repository and find your subnet IDs:

AWS Console

Create ECR repository:

1. Open Amazon ECR
2. Click Create repository
3. Set repository name to my-agentos
4. Keep defaults and click Create repository
5. Copy the repository URI (e.g., [ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com/my-agentos)

Authenticate Docker with ECR:

Click View push commands in the ECR console and run the login command:

aws ecr get-login-password --region us-east-1 | \
  docker login --username AWS --password-stdin \
  [ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com

ECR tokens expire after 12 hours. Re-run this command if you get authentication errors.

Find your subnet IDs:

1. Open VPC Console → Subnets
2. Look for subnets with Auto-assign public IP = Yes
3. Note the Subnet IDs for 2+ subnets in different availability zones

Public subnets typically have “public” in their name or have a route to an Internet Gateway.

CLI

Create ECR repository:

aws ecr create-repository \
  --repository-name my-agentos \
  --region us-east-1

Note your ECR URL from the output:

[ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com

Authenticate Docker with ECR:

aws ecr get-login-password --region us-east-1 | \
  docker login --username AWS --password-stdin \
  [ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com

ECR tokens expire after 12 hours. Re-run this command if you get authentication errors.

Find your subnet IDs:

aws ec2 describe-subnets --query 'Subnets[*].[SubnetId,AvailabilityZone,MapPublicIpOnLaunch]' --output table

Choose subnets where MapPublicIpOnLaunch is True. Select subnets in different availability zones for high availability.

Step 2: Configure

1. Update settings

   Edit infra/settings.py with your AWS details:

   infra_settings = InfraSettings(
       infra_name="my-agentos",
       aws_region="us-east-1",
       aws_subnet_ids=["subnet-xxx", "subnet-yyy"],  # Your subnet IDs
       image_repo="[ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com",
       push_images=True,
   )
   
2. Create secrets files

   cp infra/secrets/prd_api_secrets.example.yml infra/secrets/prd_api_secrets.yml
   cp infra/secrets/prd_db_secrets.example.yml infra/secrets/prd_db_secrets.yml
   

   Edit both files with your credentials. See Secrets for details.

Step 3: Test & Deploy

1. Test locally (optional)

   ag infra up
   

   Open http://localhost:8000/docs to verify your agents load correctly.

2. Deploy to AWS

   ag infra up prd:aws
   

   Press Enter to confirm. This creates:

   • Docker image pushed to ECR
   • RDS PostgreSQL instance with pgvector
   • ECS Cluster, Service, and Task Definition
   • Application Load Balancer with Target Group
   • Security Groups for network isolation

   RDS takes 5-10 minutes to provision. The deployment will wait for it automatically.

Step 4: Get Your Endpoint

AWS Console

1. Open EC2 Console → Load Balancers
2. Find your load balancer (named my-agentos-...)
3. Copy the DNS name

Test your deployment:

curl http://[LOAD_BALANCER_DNS]/health

CLI

aws elbv2 describe-load-balancers \
  --query 'LoadBalancers[?contains(LoadBalancerName, `my-agentos`)].DNSName' \
  --output text

Test your deployment:

curl http://[LOAD_BALANCER_DNS]/health

Should return: {"status": "ok", "instantiated_at": "..."}

---

Next Steps

Configure Settings

Region, subnets, and ECR

Set Up Secrets

API keys and credentials

Verify Deployment

Confirm everything is working

---

Troubleshooting

'No basic auth credentials' error

ECR tokens expire after 12 hours. Re-run the authentication command:

aws ecr get-login-password --region us-east-1 | \
  docker login --username AWS --password-stdin \
  [ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com

RDS provisioning takes too long

RDS typically takes 5-10 minutes. Check status in RDS Console → Databases. Look for the database status to change from “Creating” to “Available”.

Can't find public subnets

Public subnets have a route to an Internet Gateway. In the VPC Console:

1. Go to Route Tables
2. Find tables with a route to igw-xxx (Internet Gateway)
3. Check which subnets are associated with those route tables

Health check returns 502 Bad Gateway

The container may still be starting. Wait 2-3 minutes and retry.

If it persists, check ECS task logs:

1. Open ECS Console → Clusters → my-agentos
2. Click on the running task
3. Go to Logs tab to see container output

Task keeps stopping and restarting

This usually indicates the container is crashing. Common causes:

• Missing environment variables (check your secrets files)
• Invalid API keys
• Database connection issues

Check the Stopped tasks in ECS Console for error messages.