AWS re:Invent 2025 - Accelerate Terraform Provider development workflows with Kiro (DVT216)

Accelerating Terraform Provider Development with Agentic AI

The Challenge: Scaling Terraform Provider Development at HashiCorp

• HashiCorp's AWS Terraform provider has seen exponential growth, with over 3,000 open issues, 518 open PRs, and 3,000+ contributors
• The single HashiCorp engineering team struggles to keep up with the increasing number of changes, issues, and contributions
• Scaling the team is difficult due to the specialized Terraform, Golang, and AWS expertise required
• This leads to a review bottleneck, with limited expert reviewers available to ensure code quality and security

Exploring AI Solutions

• HashiCorp initially explored using raw language models, but found the results unpredictable and hard to repeat
• Instruction-based agents provided some guidance, but still required significant user oversight and could drift off course
• Speculative agents with Kira have delivered consistent, high-quality work for HashiCorp

Integrating Agentic AI into the SDLC

Analysis and Planning

• HashiCorp uses PRFAQs, PRDs, and RFCs to capture requirements and design decisions
• Kira assists in producing these documents, refining them iteratively with the engineering team

Design and Implementation

• Kira generates detailed implementation plans based on the requirements, including step-by-step code development
• The engineering team collaborates with Kira, providing feedback and ensuring the code meets quality standards

Testing and Integration

• Kira writes unit and end-to-end tests for the new Terraform resources and actions
• The team reviews the test results and makes any necessary adjustments to the implementation

Documentation

• Kira automatically generates documentation for the new Terraform resources, following HashiCorp's standard formats

Results and Benefits

• 90% faster development time for new Terraform resources and actions
• Over 700 high-quality configuration documentation resources generated
• Significant time savings across the SDLC, freeing up engineers to focus on higher-value tasks

HashiCorp's MCP Server for AI-Powered Workflows

• HashiCorp has developed an MCP (Model Coordination Platform) server to integrate AI agents into their Terraform workflows
• The MCP server allows users to quickly set up AI-powered assistance for common tasks, such as creating S3 buckets with Terraform
• The MCP server pulls from multiple sources to provide comprehensive, policy-compliant solutions

The Future: Deeper Human-Agent Collaboration

• Agents are gaining greater autonomy, enabling "organic transformation" use cases
• True collaboration between humans and agents, with effective oversight, is key to high-quality output
• Agentic AI is transforming the SDLC, enabling more inclusive and efficient software development