Deployment options

Run Argent where your data lives

Argent Strategic Intelligence installs inside an environment you control: on your own servers with no internet, in your private cloud, or in your own AWS, Azure or Google Cloud account. One signed release, the same controls, three ways to run it.

01 / DEPLOYMENT OPTIONS

One product, three places to run it

Argent is the gateway between your AI agents and the AI models they use. It checks each request, records it on a signed record, and stops what your policy forbids. In every option below, that gateway runs inside your environment, and so do its records.

Where AI runs is now a buying question in its own right. Gartner forecasts worldwide sovereign cloud infrastructure spending of $80 billion in 2026, up 35.6% from 2025 (Gartner, February 2026).

  • [Air-gapped on-premises](/platform/deployment/air-gapped-on-prem/): your own servers, a network with no internet. The release is carried in on removable media.
  • [Private cloud](/platform/deployment/private-cloud/): infrastructure only your organization uses, in your data center or a dedicated cloud, connected where you allow it.
  • [Public cloud VPC](/platform/deployment/public-cloud-vpc/): your own AWS, Azure or Google Cloud account, inside a network you define, with the cloud's managed services if you want them.
02 / DEPLOYMENT OPTIONS

The three options side by side

Pick the column that matches your network. The checks Argent runs on each request are the same in all three.

Air-gapped on-premisesPrivate cloudPublic cloud VPC
Where it runsYour servers, in a sealed networkYour data center or a dedicated cloudYour AWS, Azure or Google Cloud account
InternetNoneWhere you allow itWhere you allow it
How the release arrivesOne signed bundle on removable media, checked and loaded into your registryFrom the Nyedis registry with signatures checked, or from your own mirrorFrom the Nyedis registry, or mirrored into your cloud registry
Install withHelm chart (self-hosted profile) or the single-server bundleHelm chart (self-hosted profile) or the single-server bundleHelm chart with a profile for Amazon EKS, Azure AKS or Google GKE
Database, cache, secrets, evidenceAll inside your installAll inside your installIn your cluster, or your cloud's key service and locked storage (plus Aurora and ElastiCache on AWS)
AI modelsLocal engines such as Ollama, LM Studio and vLLMLocal engines, plus outside providers you approveCloud providers such as Amazon Bedrock, Azure OpenAI and Google Gemini, plus any you approve
UpdatesThe next signed bundle, by mediaThe next signed releaseThe next signed release
03 / DEPLOYMENT OPTIONS

What is the same in every option

Every way of running Argent is built from one signed release manifest, so the image digests agree and no form factor ships different bytes.

  • The same gateway. It knows which agent sent each request and which person owns that agent, and checks every request that works through Argent against your policy.
  • The same signed records. Each decision is written to a signed, tamper-evident record, and the archive is kept on object-locked storage you choose.
  • The same console. It loads only from your install.
  • The same rollout. Start in watch mode, where Argent records what it would have done while nothing is stopped, then turn on enforcement area by area.
  • No provider keys in agents. The gateway holds the real keys, resolved from your vault, and agents carry only an Argent key, in a public cloud they can carry none at all and sign in with their own cloud identity. See keyless agents.
  • Two doors. On Kubernetes only the gateway and the console face the network; the chart's network policies keep everything else inside the cluster, where the cluster's network plugin enforces them.
  • Evidence anyone can check. An auditor checks the evidence package on their own machine, with no call to Nyedis. See one-click audit evidence.
04 / DEPLOYMENT OPTIONS

How to choose

Four plain questions settle it.

If your answer is...Start with
The network has no internet, by designAir-gapped on-premises
We run our own Kubernetes or servers, and keep data in our own data centerPrivate cloud
Our workloads and AI models already live in AWS, Azure or Google CloudPublic cloud VPC
We want a quick pilot on one machine firstThe single-server bundle, in any of the three

Every option installs the same signed release, so a pilot on one machine and a production cluster run the same images.

05 / DEPLOYMENT OPTIONS

How long an install takes

These are measured runs, start to every service ready.

WhereMeasured
AWS, fresh accountEleven runs took 17.7 to 35.1 minutes, a median of 28.9
AzureEvery service ready at 16.5 minutes (one run)
Google Cloud21 and 23.4 minutes from creating the project to ready (two runs)
One serverServing 100 seconds after unpacking, with images already in a private registry

An air-gapped install adds the time it takes to carry in and load the bundle.