Using the Agent

Once in the workspace, getting started is as simple as expressing what you want to build in natural language. The agent works with you to understand your intent, plan the approach, and generate the infrastructure code.

Behind the scenes, the agent reads your files, references your MCP servers, follows your rulesets, delegates to subagents, and executes your workflows to work the way your organization works.

How the agent interacts with you depends on your personal agent preferences. You can set your experience level and working style so the agent adapts to how you like to work.

The workspace layout

The workspace has a chat panel on one side and a content area on the other. The content area can show different views depending on what you're working on. It can display your code editor for browsing and editing files, diagrams generated by the agent, or diffs from your history. Type a message in the chat and the agent responds, creating files, modifying code, answering questions, or explaining concepts.

What you can do with it

• Create new infrastructure from scratch using Terraform, OpenTofu, Pulumi, and other infrastructure as code languages
• Bring existing cloud resources under IaC management by generating code that represents what's already running in your environment
• Migrate between clouds or build multi-cloud ecosystems by translating infrastructure patterns across providers
• Modernize on-prem systems by translating requirements from existing environments into cloud infrastructure as part of transformation activities
• Audit and review existing infrastructure code for security, cost, and best practice improvements
• Create reusable modules that embed your organization's standards and can be shared across workspaces
• Generate documentation for your infrastructure, architecture decisions, and operational procedures
• Create diagrams including architecture diagrams at different levels of detail, physical cloud architecture, threat vector maps, and more. You can create multiple diagrams per workspace and toggle between them from the diagram panel.

Conversations

Each conversation is a persistent thread. The agent remembers everything discussed, has access to all workspace files and docs, and considers your rulesets throughout.

You can have multiple conversations in a workspace. Useful for working on different features or keeping focused threads.

Tips

• Be specific. "Create a Terraform module for an S3 bucket with versioning and encryption" works better than "Make an S3 bucket"
• Reference files. Mention specific files to focus the agent's attention, or use @ to reference uploaded docs
• Iterate. Build on the agent's output step by step
• Use workflows. For complex tasks, run a workflow to guide the agent through a structured process
• Connect MCP servers. Add MCP servers to give the agent context from existing systems like Terraform Cloud or AWS, or let it take action in external platforms
• Upload docs. Give the agent project-specific context like requirements, architecture plans, and design specs
• Add rulesets. Define standards and guidelines so the agent follows your organization's way of working
• Start with context. Explain what you're working on and what you need