# AgentCore Deployment - Final Steps
## Current Status
Your configuration is **almost ready**! The toolkit is having trouble with the `ecr: auto` setting. Here are two solutions:
## ✅ Solution 1: Create ECR Repository Manually (RECOMMENDED)
Run this on your EC2 instance:
```bash
# Create ECR repository
./create-ecr-repo.sh
# Get the repository URI
REPO_URI=$(aws ecr describe-repositories --repository-names aws-pillar-mcp-server --region us-east-1 --query 'repositories[0].repositoryUri' --output text)
echo "Repository URI: $REPO_URI"
```
Then update `.bedrock_agentcore.yaml` to use the repository URI instead of `auto`:
```yaml
agents:
aws-finops-mcp:
name: aws-pillar-mcp-server
description: AWS FinOps MCP Server for cost optimization and resource management
entrypoint: src/aws_finops_mcp/__main__.py
deployment_type: container
aws:
account: null
region: us-east-1
execution_role: auto
ecr: aws-pillar-mcp-server # Use repository name instead of 'auto'
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
```
Then deploy:
```bash
agentcore launch
```
## ✅ Solution 2: Use CLI Configure Command
Instead of manually editing the YAML, use the CLI to configure everything:
```bash
# Configure the agent with all required settings
agentcore configure \
--entrypoint src/aws_finops_mcp/__main__.py \
--name aws-pillar-mcp-server \
--deployment-type container \
--ecr auto \
--execution-role auto \
--region us-east-1 \
--protocol MCP \
--non-interactive
# Then deploy
agentcore launch
```
## 🔍 Understanding the Error
The error `ECR repository not configured and auto-create not enabled` suggests that:
1. The toolkit might not support `ecr: auto` in YAML format
2. OR it needs to be configured via CLI first
3. OR there's a version mismatch in the toolkit
## 📋 Complete Deployment Steps
### Step 1: Create ECR Repository
```bash
# On your EC2 instance
cd /root/aws-pillar-mcp-server
# Create the repository
aws ecr create-repository \
--repository-name aws-pillar-mcp-server \
--region us-east-1 \
--image-scanning-configuration scanOnPush=true \
--encryption-configuration encryptionType=AES256
# Verify it was created
aws ecr describe-repositories \
--repository-names aws-pillar-mcp-server \
--region us-east-1
```
### Step 2: Update Configuration
Edit `.bedrock_agentcore.yaml` and change:
```yaml
ecr: auto
```
To:
```yaml
ecr: aws-pillar-mcp-server
```
### Step 3: Deploy
```bash
agentcore launch
```
## 🎯 Expected Deployment Flow
Once the ECR issue is resolved, you should see:
```
🚀 Launching Bedrock AgentCore (codebuild mode - RECOMMENDED)...
Memory disabled - skipping memory creationStarting CodeBuild ARM64 deployment for agent 'aws-pillar-mcp-server' to account 058029412961 (us-east-1)
Generated image tag: 20260129-XXXXXX-XXX
Setting up AWS resources (ECR repository, execution roles)...
✅ ECR repository configured: aws-pillar-mcp-server
✅ Creating execution role...
✅ Starting CodeBuild project...
⏳ Building container image (this takes 5-10 minutes)...
✅ Container image built and pushed to ECR
✅ Creating AgentCore Runtime...
✅ Deploying agent...
✅ Agent deployed successfully!
Agent ARN: arn:aws:bedrock-agentcore:us-east-1:058029412961:agent/XXXXX
Agent URL: https://XXXXX.execute-api.us-east-1.amazonaws.com
Test with:
agentcore invoke '{"prompt": "list unused resources"}'
```
## 🆘 Alternative: Manual Docker Deployment
If `agentcore launch` continues to have issues, you can deploy manually:
```bash
# Set variables
export AWS_REGION=us-east-1
export AWS_ACCOUNT_ID=058029412961
# Create ECR repository (if not exists)
aws ecr create-repository --repository-name aws-pillar-mcp-server --region $AWS_REGION || true
# Build for ARM64
docker buildx build --platform linux/arm64 -t aws-finops-mcp:latest -f Dockerfile.agentcore .
# Login 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
# Tag and push
docker tag aws-finops-mcp:latest \
$AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com/aws-pillar-mcp-server:latest
docker push $AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com/aws-pillar-mcp-server:latest
# Deploy using Python
python3 << 'EOF'
from bedrock_agentcore_starter_toolkit import Runtime
import os
runtime = Runtime()
runtime.configure(
image_uri=f"{os.getenv('AWS_ACCOUNT_ID')}.dkr.ecr.{os.getenv('AWS_REGION')}.amazonaws.com/aws-pillar-mcp-server:latest",
env_vars={
'AWS_REGION': os.getenv('AWS_REGION'),
'MCP_SERVER_MODE': 'http',
'MCP_SERVER_HOST': '0.0.0.0',
'MCP_SERVER_PORT': '8000'
}
)
result = runtime.launch(local=False)
print(f"✅ Agent deployed!")
print(f"Agent ARN: {result.agent_arn}")
EOF
```
## 📊 After Successful Deployment
### Test Your Agent
```bash
# Basic test
agentcore invoke '{"prompt": "hello, what can you do?"}'
# Cost optimization test
agentcore invoke '{"prompt": "find unused EC2 instances in us-east-1"}'
# Security audit test
agentcore invoke '{"prompt": "check for security vulnerabilities"}'
```
### Monitor Logs
```bash
# View logs
agentcore logs --follow
# Or use CloudWatch
aws logs tail /aws/bedrock-agentcore/runtime/aws-pillar-mcp-server --follow
```
### Check Status
```bash
agentcore status
```
## 🔧 Troubleshooting
### Issue: "ECR repository not configured"
**Solution**: Create the ECR repository manually first (see Step 1 above)
### Issue: "execution_role not configured"
**Solution**: The `execution_role: auto` should work, but if not, create manually:
```bash
# Create execution role
aws iam create-role \
--role-name bedrock-agentcore-execution-role \
--assume-role-policy-document '{
"Version": "2012-10-17",
"Statement": [{
"Effect": "Allow",
"Principal": {"Service": "bedrock-agentcore.amazonaws.com"},
"Action": "sts:AssumeRole"
}]
}'
# Attach policies
aws iam attach-role-policy \
--role-name bedrock-agentcore-execution-role \
--policy-arn arn:aws:iam::aws:policy/AmazonBedrockFullAccess
# Update config with role ARN
```
### Issue: CodeBuild fails
**Solution**: Check CodeBuild logs:
```bash
aws codebuild list-builds --region us-east-1
aws codebuild batch-get-builds --ids <build-id> --region us-east-1
```
### Issue: Container fails health check
**Solution**: Test locally first:
```bash
# Build and run locally
docker build -t test -f Dockerfile.agentcore .
docker run -p 8000:8000 -e MCP_SERVER_MODE=http test
# Test health endpoint
curl http://localhost:8000/health
```
## 📚 Next Steps After Deployment
1. **Test all tool categories**:
```bash
agentcore invoke '{"prompt": "list all available tools"}'
```
2. **Set up monitoring**:
- CloudWatch dashboards
- Log insights
- Alarms for errors
3. **Configure authentication** (optional):
- OAuth with Cognito
- API keys
- IAM authentication
4. **Optimize costs**:
- Monitor usage
- Adjust idle timeout
- Use reserved capacity
## 🎓 Summary
The key issue is that `ecr: auto` might not be working as expected. The simplest solution is:
1. **Create ECR repository manually**: `./create-ecr-repo.sh`
2. **Update config**: Change `ecr: auto` to `ecr: aws-pillar-mcp-server`
3. **Deploy**: `agentcore launch`
This should get your agent deployed successfully!
---
**Need help?** Check the other guides:
- `MANUAL_AGENTCORE_DEPLOY.md` - Detailed manual deployment
- `AGENTCORE_RUNTIME_REVIEW.md` - Status and troubleshooting
- `QUICK_FIX_DEPLOYMENT.md` - Quick deployment guide
