
Argent on Kubernetes with Helm
One Helm chart installs Argent Strategic Intelligence on any Kubernetes 1.34 or newer: Amazon EKS, Azure AKS, Google GKE, k3s or RKE2. Every image comes from the same signed release, pinned by digest.
What Argent does with Kubernetes
Most teams that run containers already run Kubernetes in production. The CNCF's 2025 survey found that "82% of container users now run Kubernetes in production, up from 66% in 2023." Source: CNCF.
Argent Strategic Intelligence is the gateway between your AI agents and the AI models they use. On Kubernetes it installs from one Helm chart, the one supported install path for production.
- One chart for every substrate. Amazon EKS, Azure AKS, Google GKE, k3s and RKE2, on Kubernetes 1.34 or newer.
- The same bytes everywhere. The chart, the single-server bundle and the air-gapped bundle are all made from one signed release, so the image digests agree.
- Only two doors face the network: the gateway and the console. The chart's network policies keep everything else inside the cluster, where the cluster's network plugin enforces them.
What the chart installs
Everything the gateway needs, in one release:
| Part | What it does |
|---|---|
| The gateway | Every agent request lands here. It decides, masks sensitive data and records each call, with the sensitive-data detector beside it. |
| The edge | The Envoy proxy every request enters through, scaled out as traffic grows. |
| The control plane and console | Policy, agents, owners, budgets and the operator console. |
| Upgrade jobs | The database migrates and its indexes build in their own jobs, before the app starts. |
| The offline verifier | Checks the signed records independently, with every standard install. |
| The data tier | Database, cache, secrets store and object-locked evidence store, in the cluster or as your cloud's managed services. |
| Monitoring | Prometheus, Alertmanager, Grafana, Loki, Tempo and OpenTelemetry, with retention settings. See Prometheus and Grafana. |
In the cluster, or managed. PostgreSQL through CloudNativePG or Amazon Aurora; Valkey or Amazon ElastiCache; OpenBao in the cluster or unlocked by your cloud's key service; an object-locked store in the cluster or Amazon S3 Object Lock in compliance mode, an Azure Blob container with version-level immutability, or Google Cloud Storage.
A profile for each place it runs
Pick a profile, then a size.
| Profile | What it sets up |
|---|---|
| AWS (EKS) | Exposes only the gateway and the console, through AWS load balancers. See Amazon Web Services. |
| Azure (AKS) | The gateway's edge handles TLS itself; the secrets store unlocks from Azure Key Vault through workload identity. See Microsoft Azure. |
| Google Cloud (GKE) | The gateway's edge handles TLS itself; the secrets store unlocks from Google Cloud KMS through workload identity. See Google Cloud. |
| Self-hosted (k3s, RKE2) | Runs the database, cache, secrets store and evidence store inside your cluster, for private cloud and air-gapped sites. |
| Production | Keeps at least three gateway copies running. |
Four sizes. Sizing profiles set resources, storage and default rate limits. Default rate limits are only seeded where none exist, so an operator's own limits are never overwritten.
What you set in the chart
A handful of values carry the decisions that are yours to make:
| Value | What it decides |
|---|---|
global.imageRegistry | Swaps the registry host of every image and keeps each digest, so the same signed release runs from an Amazon ECR mirror, your own mirror or an air-gapped registry. |
objectStore.backend | Where the evidence archive lives: s3 (AWS S3, MinIO, SeaweedFS), azure or gcs. |
objectStore.lock | How long signed records stay locked, seven years by default. Argent refuses to start with a weaker lock unless an operator allows it. |
gateway.audit.overflow | If the record buffer fills: refuse with a 503 (the default), or serve with signed gap markers by operator choice. |
gateway.decisionAPI.enabled | Off by default. Turns on the decision listener for agent hooks. |
connectorEgressAllowlist | Under controlPlane. Makes connector traffic deny-by-default, so only listed addresses are reachable. |
tenantEgressPrivateCidrs | Under controlPlane. Re-opens private ranges for a tenant's own vault, MCP server or directory. Cloud metadata addresses stay refused. |
Built to stay up, and to fail safe
The chart's defaults favour a refused request over an unrecorded one.
- Grows with demand. Optional autoscaling on CPU, open streams, sensitive-data queue pressure and record buffer depth: the signals that come before a refusal.
- High availability by default. Disruption budgets, topology spread and separate service accounts.
- No drift from the release. The chart refuses to render if an image is missing from the release, and a conformance check proves a rendered chart installs exactly the release's images, each once.
- Clean upgrades. On shutdown a pod stops taking new work, finishes what is in flight and flushes its records. A cut stream is refused, never passed unchecked.
Measured in real cloud accounts. On a fresh AWS account, eleven installs took 17.7 to 35.1 minutes, a median of 28.9. Azure was ready at 16.5 minutes (one run), and Google Cloud at 21 and 23.4 minutes (two runs). In four production-profile upgrade runs on AWS, between 98,582 and 164,100 gateway requests passed through each upgrade, with 0 failed requests caused by the upgrade.
Where it fits, and how you start
The chart is how Argent runs in production, in a cloud or in your own data centre.
- Where Argent runs: air-gapped on premises, private cloud or your public cloud VPC. Compare the options.
- Capabilities: air-gapped install, shadow AI discovery, model control, keyless agents, watch before you enforce and one-click audit evidence. All capabilities.
- Goals: see all AI in your company, model control and routing, threat defense, compliance evidence and sovereign AI. All goals.
- Use cases: AI coding agents, customer-facing assistants, enterprise copilots, agent tool access and regulated and air-gapped AI.
- Related integrations: HashiCorp Vault and OpenBao, Prometheus and Grafana, OpenTelemetry, Tempo and Loki, Docker Compose and the air-gapped bundle.
- Industries: financial services, insurance, healthcare, the public sector and retail. All industries.
- Install. Pick your profile and size, choose in-cluster or managed data services, and install the chart.
- Route. Point your agents at the gateway, by key or by their cloud identity, and your routes at the models you approve.
- Watch, then enforce. Start in Watch mode, then switch on Enforce. Take the walkthrough.
Want this running on your own use cases?
Book a demo