Skip to main content

​
What It Does

Type what you want in plain English. Kyrna’s AI generates Terraform code. Status: Working (MVP)

​
How It Works

1

You Describe

Type what infrastructure you need in natural language
2

AI Generates

Claude 3.5 Sonnet or GPT-4 creates Terraform code
3

You Review

See generated code, costs, and security issues
4

You Deploy

Approve and deploy

​
AI Models

Claude 3.5 Sonnet (recommended)
  • Better for infrastructure code
  • More accurate resource configs
  • Default model
GPT-4 (alternative)
  • Also works well
  • Fallback option
Status: Both working

​
What It Can Generate

AWS Resources:
  • EC2 instances
  • RDS databases
  • S3 buckets
  • VPC networking
  • Lambda functions
  • DynamoDB tables
  • ELB/ALB load balancers
  • Security groups
  • IAM roles/policies
  • And more…
Current Limitation: AWS + Terraform only

​
Example

Simple: 
You: "Create a web server"

Kyrna generates:
- aws_instance (EC2)
- aws_security_group (SSH + HTTP)
- aws_eip (public IP)
Complex: 
You: "Create a production web app with database and load balancer"

Kyrna generates:
- VPC with subnets
- Application Load Balancer
- Auto Scaling Group
- RDS PostgreSQL (multi-AZ)
- Security groups
- IAM roles
- All configurations

​
Current Capabilities

✅ Multi-resource generation (multiple resources at once)
✅ Context-aware (understands relationships)
✅ Best practices (secure defaults)
✅ Cost-optimized (reasonable instance sizes)

​
Current Limitations

❌ AWS only (no GCP, Azure, DigitalOcean)
❌ Terraform only (no Pulumi, CDK, CloudFormation)
❌ English only (no other languages)
❌ May need refinement (MVP, not perfect)

​
Accuracy

What to expect:
  • Usually generates correct code
  • May need small tweaks
  • Follows Terraform best practices
  • Sometimes misses edge cases
Always review before deploying

​
What’s NOT Included

  • Visual drag-and-drop
  • Template library
  • Code completion (coming)
  • Multi-language support

​
Next Steps

Try It

Generate your first infrastructure

Examples

See more examples