Can Topsoil write Terraform or CloudFormation?
Yes, as repository files. It plans, edits in a worktree and opens a pull request for a person to review. Applying the stack stays on your existing deploy path.
Learn
Software, Infrastructure and Pipeline Work on Amazon Bedrock
Topsoil is a private AI platform for engineering. This guide is the work it takes on: application changes, infrastructure as code and pipeline files, on Amazon Bedrock in your own AWS account.
Application Source
A backlog item becomes a proposed change in stages: explore, design, plan, implement, verify, review, document. You approve the plan before files change. Implementation happens in an isolated git worktree on its own branch. New tests have to fail with the rest of the change put back before verify can pass, for the languages and runners Topsoil supports, or for commands you declare in .topsoil/tests.json. A second model can review the diff. A person merges.
Plan approval, diffs and the command allowlist in that sequence are written out in how engineering changes get approved.
Infrastructure as Code
Terraform, CloudFormation templates, and the policy and module files beside them are ordinary repository files. Topsoil can plan a change, edit the files in a worktree and open a pull request. A command such as plan or validate runs only when it matches your allowlist. Applying a stack, or changing a live resource, stays with the review and deploy path you already run.
Pipelines and DevSecOps Files
Pipeline definitions, workflow YAML, Dockerfiles and policy-as-code live in the same repositories. Topsoil proposes edits under the same plan approval. On a forge that supports it, Topsoil can read CI results on the pull request. On self-managed GitLab or Bitbucket Data Center, a CI job can post a /review as a comment. An unattended run needs a budget ceiling. A budget stop exits with its own code, so a scheduler can tell it from a failure. Budgeting on Amazon Bedrock names that code.
Your CI system still runs the jobs. Topsoil edits the definitions in git and reads results where the forge allows. Vulnerability scanning stays with the tools you already use for it.
Model Choice and Account Setup
Which model runs, which Region it runs in, and whether the path is public Bedrock, custom HTTPS or PrivateLink are decisions for your account. Those steps are written on Amazon Bedrock, deployment and AWS GovCloud (US).
Government and defense teams comparing a vendor-hosted assistant have their own page: a Copilot alternative for government and defense engineering.
Questions
Frequently asked questions
Does it run our CI pipelines?
Your CI system runs the jobs. Topsoil can edit pipeline files in git and read CI results on pull requests where the forge supports that.
Is this only for application code?
Design documents, software, infrastructure as code and CI configuration are all in scope when they live in a project Topsoil can open.
Do we need GPUs?
No. Inference uses Amazon Bedrock models in your account.
Try it on your own codebase
A pilot runs in your own AWS account, with your models and your material, and ends with findings you can act on.