Plan a phased cloud migration with checkpoints

Use an agent to break a large migration into approved phases and carry out each step test-first.

Done with Gemini· Advanced· Apr 2026

What they did
A developer ran Gemini CLI with the Conductor extension to manage a multi-step cloud migration. They kept product, stack, and workflow details in Markdown files, and for each track the agent drafted a spec and execution plan before the developer signed off. After approval, it worked test-first, wrote Terraform, and paused at checkpoints at the end of each phase.
How it went
The migration took several days of AI-assisted work. Usage came to about 19M input tokens, 66M cached input tokens, and 400k output tokens, for roughly $30 on Vertex AI. Cached input dominated the bill, and the source notes that prices can change.
Tips
Keep the agent in a sandbox and require approval before any irreversible step. Put procedural rules such as running `terraform plan` before `apply` and your coverage target in a workflow file.

Try it yourself with Claude Code

I’m migrating [current system] to [target cloud/platform]. Read the files in this repo that describe the product, stack, and workflow, then split the work into phases with checkpoints, risks, and rollback steps. For each phase, draft a short spec and plan, then stop and wait for my approval before you change code, run infrastructure commands, or make any destructive edits. Work test-first, use the repository’s workflow rules, and after each phase give me a brief checkpoint summary in [MIGRATION.md or another file].

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on dev.to ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.