amicu

amicu vs Terraform and OpenTofu

Terraform and OpenTofu are general-purpose infrastructure-as-code tools: they manage any resource that a provider plugin exposes. amicu is a much narrower open-source CLI that deploys vLLM and llama.cpp servers from HCL and understands models, memory fit and endpoint authentication out of the box. They share the write, plan, apply workflow, so the two are not rivals: use Terraform or OpenTofu for your wider infrastructure, and amicu when the thing you are managing is an LLM endpoint.

QuickstartAll comparisons

Side by side

amicu vs Terraform and OpenTofu
 Terraform and OpenTofuamicu
ScopeAny infrastructure with a provider: clouds, Kubernetes, SaaS and more.Open-weight LLM inference only: models, hardware profiles and deployments.
WorkflowWrite, plan, apply, with state as the record of real infrastructure.validate, plan, apply, destroy, with state recording what was applied.
Configuration languageDeclarative configuration files with reusable modules.HCL (amicu.hcl) with a fixed set of resource kinds such as model, hardware_profile and deployment.
Knows what a model needsNo built-in concept of a model or of GPU memory fit.validate checks that model weights plus KV cache fit the accelerator or host memory.
Serving enginesNone built in; you describe the machines or containers yourself.Starts vLLM or llama.cpp for you and exposes its OpenAI-compatible server.
Endpoint checksNot part of the core workflow.After apply, verifies that the endpoint rejects anonymous requests.
Where the workload runsWherever a provider and your configuration point.This machine, or GPU cloud providers such as Modal and RunPod.
State storageA state file as the source of truth for your environment; remote backends such as S3, GCS and Azure are available.Local state file in the project only; there is no remote backend yet.
Ecosystem and maturityThousands of providers and a large community.Small and early. Not stable; the HCL schema and CLI may change.
LicenseTerraform: BUSL 1.1 (since 2023). OpenTofu: MPL-2.0.AGPL-3.0-only.

Sources for the Terraform and OpenTofu column: Terraform introduction, OpenTofu documentation. Details can change; check the official documentation.

When to choose Terraform and OpenTofu

  • You manage networks, databases, clusters and more, and want one tool for all of it.
  • You need remote state, modules and the provider ecosystem your team already uses.
  • Your LLM serving already runs on Kubernetes or VMs that you define yourself.

When to choose amicu

  • The resource you care about is an LLM server, and you want fit checks before anything starts.
  • You want to deploy vLLM or llama.cpp to your own machine or a GPU provider without writing the plumbing.
  • You want residency and provider allowlist rules checked by validate and plan.

Questions

Can I use amicu and Terraform together?
Yes. They manage different things. Terraform or OpenTofu can provision the rest of your infrastructure while amicu deploys and tracks the LLM servers. amicu does not read Terraform state.
Is amicu a Terraform provider?
No. amicu is a standalone command-line tool with its own HCL schema and its own state file.
Does amicu support remote state like Terraform?
Not yet. State is a local file in the project directory, protected by a local lock. Remote state backends are not implemented.
Is amicu production-ready?
amicu is early and not stable yet. Expect breaking changes to the HCL schema and the CLI.

Try amicu

Install the CLI and deploy a model on your machine or a GPU provider.

QuickstartProviders