Open-weight LLMs, deployed as code.

On your laptop, your servers or a cloud GPU.

terminal

$ amicu apply
Plan: 1 action(s).
 
  + deployment.chat
      action:   create
      hardware: gpu.l4
      target:   modal (region eu-west)
      api_key:  ****** (secret "CHAT_API_KEY", resolved at apply)
Apply these changes? [y/N] y
 
Applied 1 deployment(s):
  deployment.chat -> https://<workspace>--amicu-chat-server.<region>.modal.direct
$ curl 'https://<workspace>--amicu-chat-server.<region>.modal.direct/v1/chat/completions' \
  -H "Authorization: Bearer $CHAT_API_KEY" -H "Modal-Key: $MODAL_PROXY_TOKEN_ID" \
  -H "Modal-Secret: $MODAL_PROXY_TOKEN_SECRET" -H "Content-Type: application/json" \
  -d '{"model": "RedHatAI/Qwen3.5-9B-FP8-dynamic", "messages": [{"role": "user", "content": "Hello"}]}'
{
"id": "chatcmpl-8f3a1c92d7e4",
"object": "chat.completion",
"created": 1767225600,
"model": "RedHatAI/Qwen3.5-9B-FP8-dynamic",
"choices": [{"index": 0, "message": {"role": "assistant", "content": "Hello! How can I help you today?"}, "finish_reason": "stop"}],
"usage": {"prompt_tokens": 9, "completion_tokens": 10, "total_tokens": 19}
}

$ curl 'https://<workspace>--amicu-chat-server.<region>.modal.direct/v1/chat/completions' \ -H "Authorization: Bearer $CHAT_API_KEY" -H "Modal-Key: $MODAL_PROXY_TOKEN_ID" \ -H "Modal-Secret: $MODAL_PROXY_TOKEN_SECRET" -H "Content-Type: application/json" \ -d '{"model": "RedHatAI/Qwen3.5-9B-FP8-dynamic", "messages": [{"role": "user", "content": "Hello"}]}'{ "id": "chatcmpl-8f3a1c92d7e4", "object": "chat.completion", "created": 1767225600, "model": "RedHatAI/Qwen3.5-9B-FP8-dynamic", "choices": [{"index": 0, "message": {"role": "assistant", "content": "Hello! How can I help you today?"}, "finish_reason": "stop"}], "usage": {"prompt_tokens": 9, "completion_tokens": 10, "total_tokens": 19} }

Describe a model. Get an endpoint.

amicu.hcl

resource "hardware_profile" "gpu" {
  provider = "modal"
  region   = "eu-west"
  accelerator {
    kind   = "gpu"
    model  = "l4"
    count  = 1
    memory = "24GiB"
  }
  memory = "32GiB"
}

resource "deployment" "chat" {
  model    = model.qwen.id
  hardware = hardware_profile.gpu.id
  runtime  = "vllm"
  api_key  = secret("CHAT_API_KEY")
  vllm {
    context_length = 200000
    port           = 8000
  }
}

terminal

$ amicu applyApply these changes? [y/N] yApplied 1 deployment(s):  deployment.chat -> https://<workspace>--amicu-chat-server.<region>.modal.direct

One file declares model, hardware and policy. amicu plans the rest.

Features

Laptop or cloud GPU. Same file.

amicu.hcl

  resource "hardware_profile" "main" {-   provider = "local"  # your machine+   provider = "modal"  # cloud GPU (Modal)+   region   = "eu-west"    ...  }

terminal

$ amicu planPlan: 1 action(s).   + deployment.chat      action:   create      hardware: gpu.l4      target:   modal (region eu-west)

Change the provider line to switch targets: your machine or a cloud GPU (Modal).

Features

Your data stays where you say.

amicu.hcl

resource "data_class" "customer" {
  residency = ["eu"]
}

resource "governance_rule" "eu_only" {
  effect          = "allow"
  allowed_regions = ["eu", "eu-west"]
}

resource "hardware_profile" "gpu" {
  provider = "modal"
  region   = "us-east"
}

terminal

$ amicu planCannot plan: configuration has errors.  error [AMICU-POLICY-002] region 'us-east' (us) violates the data residency allowlist  error [AMICU-POLICY-002] region 'us-east' violates the allowed-regions list

Residency and governance are checked at plan time, before anything deploys.

Features

Pick a provider. Change it later.

providers

Apply runs on this machine, on-prem hardware you run amicu on, and Modal today. 5 more, such as Hetzner and Scaleway, are in preview.

Features

Built for operators.

GET /v1/models
  no key    401
  with key  200

Endpoint auth, tested

After apply, amicu checks the endpoint rejects anonymous requests.

Learn more →

$ amicu skill export \
    --target claude-code

Agent skills

Teach your coding agent the amicu workflow and HCL schema.

Learn more →

$ amicu doctor
$ amicu doctor --output json

Health checks

Check the host, config and state directory in one command.

Learn more →

Plan your first deployment.

Open source, AGPL-3.0. No account needed.

Get started Providers