Does Topsoil need a Kubernetes cluster or GPUs?
No. Topsoil is a client that uses the models Amazon Bedrock serves in your account.
Compare
Topsoil and Tabnine, compared fairly.
Tabnine offers a SaaS service and, for Enterprise customers, private installations. Topsoil is a client that uses Amazon Bedrock in your own AWS account. Both aim at private AI coding; they get there differently.
The short version
Tabnine’s private installation is a Kubernetes cluster you host in your VPC or on premises, powered by open-weight models or private API endpoints. Topsoil has no server cluster: it calls the models Amazon Bedrock serves in your account.
| Topsoil | Tabnine | |
|---|---|---|
| Deployment | A client in your AWS account; no Topsoil server to install | Secure SaaS, or for Enterprise a private Kubernetes installation in a VPC (AWS, GCP or Azure) or on premises |
| Models | The Bedrock models your account enables | Tabnine’s endpoints, open-weight models you host, or private API endpoints |
| Hardware | None to run for inference | Depends on the deployment; Tabnine publishes hardware requirements for self-hosted models |
| Workflow | Staged agent runs to a pull request for a person to review | Chat, completions and agentic workflows in supported IDEs |
| Budgets | Monthly user and account budgets before each model call | Per Tabnine’s plans |
Facts about other products are taken from their public documentation, linked below. Products change; check the vendor’s current docs before deciding. If something here is out of date, tell us and we will correct it.
Which one fits
Where Tabnine is a strong fit
- You need private AI coding outside AWS, on premises or in GCP or Azure.
- You want to host open-weight models on your own hardware.
- You want completions across a broad range of IDEs.
Where Topsoil is a strong fit
- You want inference to stay in your own AWS account, on Amazon Bedrock, including AWS GovCloud (US).
- You want no vendor-operated service, prompt store or telemetry in the request path.
- You want a staged workflow (explore, design, plan, implement, verify, review, document) with a different Bedrock model per stage.
- You want monthly user and account budgets checked before each model call, with an AWS budget action as a backstop.
- Your backlog is GitHub Issues, or Jira Data Center with Bitbucket Data Center inside your own network.
The two are not mutually exclusive: teams sometimes run more than one tool for different work. Ask us about a pilot to compare on your own code.
Questions
Frequently asked questions.
See it against 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.