# Manual AgentCore Runtime Deployment Guide
Since the automated script has API compatibility issues, follow these manual steps to deploy to Amazon Bedrock AgentCore Runtime.
## Prerequisites
```bash
# Install required packages
pip install bedrock-agentcore-starter-toolkit bedrock-agentcore strands-agents boto3
# Verify installation
agentcore --help
```
## Step 1: Create AgentCore Configuration
Create `.bedrock_agentcore.yaml` in your project root:
```yaml
agents:
aws-finops-mcp:
name: aws-pillar-mcp-server
description: AWS FinOps MCP Server for cost optimization
runtime:
type: container
dockerfile: Dockerfile.agentcore
platform: linux/arm64
port: 8000
health_check:
path: /health
interval: 30
timeout: 10
retries: 3
environment:
MCP_SERVER_MODE: http
MCP_SERVER_HOST: 0.0.0.0
MCP_SERVER_PORT: "8000"
PYTHONUNBUFFERED: "1"
AWS_REGION: us-east-1
default_agent: aws-finops-mcp
```
## Step 2: Verify Dockerfile
Ensure `Dockerfile.agentcore` exists (it should already be created):
```bash
ls -la Dockerfile.agentcore
```
## Step 3: Test Locally (Optional)
```bash
# Start local development server
agentcore dev
# In another terminal, test
agentcore invoke --dev "Hello!"
```
## Step 4: Deploy to AgentCore Runtime
### Option A: Using AgentCore CLI (Recommended)
```bash
# Deploy to AWS
agentcore launch
# This will:
# 1. Build your container using AWS CodeBuild
# 2. Create necessary AWS resources
# 3. Deploy to AgentCore Runtime
# 4. Configure CloudWatch logging
```
### Option B: Using Docker + ECR + AWS CLI
If `agentcore launch` doesn't work, use this manual approach:
```bash
# 1. Set variables
export AWS_REGION=us-east-1
export AWS_ACCOUNT_ID=$(aws sts get-caller-identity --query Account --output text)
export ECR_REPO_NAME=aws-pillar-mcp-server
# 2. Build Docker image
docker buildx build --platform linux/arm64 \
-t $ECR_REPO_NAME:latest \
-f Dockerfile.agentcore .
# 3. Create ECR repository
aws ecr create-repository \
--repository-name $ECR_REPO_NAME \
--region $AWS_REGION
# 4. Authenticate Docker to ECR
aws ecr get-login-password --region $AWS_REGION | \
docker login --username AWS --password-stdin \
$AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com
# 5. Tag and push image
docker tag $ECR_REPO_NAME:latest \
$AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com/$ECR_REPO_NAME:latest
docker push $AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com/$ECR_REPO_NAME:latest
# 6. Deploy using Python SDK
python3 << 'EOF'
import boto3
from bedrock_agentcore_starter_toolkit import Runtime
import os
# Get environment variables
aws_region = os.getenv('AWS_REGION', 'us-east-1')
aws_account_id = os.getenv('AWS_ACCOUNT_ID')
ecr_repo_name = os.getenv('ECR_REPO_NAME', 'aws-pillar-mcp-server')
image_uri = f"{aws_account_id}.dkr.ecr.{aws_region}.amazonaws.com/{ecr_repo_name}:latest"
print(f"Deploying image: {image_uri}")
# Initialize Runtime
runtime = Runtime()
# Configure
runtime.configure(
image_uri=image_uri,
env_vars={
'AWS_REGION': aws_region,
'MCP_SERVER_MODE': 'http',
'MCP_SERVER_HOST': '0.0.0.0',
'MCP_SERVER_PORT': '8000'
}
)
# Launch to AgentCore Runtime
result = runtime.launch(local=False)
print(f"\n✅ Deployment successful!")
print(f"Agent ARN: {result.agent_arn}")
print(f"Agent URL: {result.agent_url}")
print(f"\nTest with: agentcore invoke '{{\"prompt\": \"list unused resources\"}}'")
EOF
```
## Step 5: Test Deployed Agent
```bash
# Test using CLI
agentcore invoke '{"prompt": "list unused lambda functions in us-east-1"}'
# Or use AWS SDK
python3 << 'EOF'
import boto3
import json
client = boto3.client('bedrock-agentcore-runtime', region_name='us-east-1')
# Replace with your agent ARN from deployment
agent_arn = "arn:aws:bedrock-agentcore:us-east-1:ACCOUNT:agent/AGENT_ID"
response = client.invoke_agent_runtime(
agentArn=agent_arn,
inputText='list unused lambda functions'
)
print(json.dumps(response, indent=2, default=str))
EOF
```
## Step 6: View Logs
```bash
# View CloudWatch logs
aws logs tail /aws/bedrock-agentcore/runtime/aws-pillar-mcp-server --follow
# Or use agentcore CLI
agentcore status
```
## Troubleshooting
### Issue: `agentcore launch` fails
**Solution**: Use Option B (Docker + ECR + AWS CLI) above
### Issue: Permission denied
**Solution**: Ensure your IAM user/role has these permissions:
```json
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": [
"bedrock-agentcore:*",
"ecr:*",
"iam:CreateRole",
"iam:AttachRolePolicy",
"iam:PassRole",
"logs:*",
"codebuild:*"
],
"Resource": "*"
}
]
}
```
### Issue: Docker build fails
**Solution**: Ensure Docker is running and you have ARM64 support:
```bash
docker buildx ls
# Should show linux/arm64 platform
```
### Issue: Image push fails
**Solution**: Re-authenticate to ECR:
```bash
aws ecr get-login-password --region us-east-1 | \
docker login --username AWS --password-stdin \
$(aws sts get-caller-identity --query Account --output text).dkr.ecr.us-east-1.amazonaws.com
```
## Alternative: AWS Console Deployment
1. **Build and push image** (Steps 1-5 from Option B above)
2. **Go to AWS Console**:
- Navigate to Amazon Bedrock
- Go to AgentCore → Runtime
- Click "Create runtime"
3. **Configure runtime**:
- Name: `aws-pillar-mcp-server`
- Image URI: `YOUR_ACCOUNT.dkr.ecr.us-east-1.amazonaws.com/aws-pillar-mcp-server:latest`
- Environment variables:
- `MCP_SERVER_MODE=http`
- `MCP_SERVER_HOST=0.0.0.0`
- `MCP_SERVER_PORT=8000`
- `AWS_REGION=us-east-1`
4. **Create and wait** for deployment to complete
5. **Test** using the console or CLI
## Next Steps
Once deployed:
1. **Monitor**: Check CloudWatch logs and metrics
2. **Test**: Invoke your agent with various prompts
3. **Optimize**: Adjust container resources if needed
4. **Scale**: Configure auto-scaling if needed
## Resources
- [AgentCore Documentation](https://docs.aws.amazon.com/bedrock-agentcore/)
- [Starter Toolkit GitHub](https://github.com/aws/bedrock-agentcore-starter-toolkit)
- [Runtime Guide](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/runtime-get-started-toolkit.html)
---
**Need Help?** See [BEDROCK_AGENTCORE_DEPLOYMENT.md](./BEDROCK_AGENTCORE_DEPLOYMENT.md) for more details.
