Install InsightStudio

Install in your environment

Three install paths, one product. Pick whichever fits your IT footprint — same code under all three.

Pick your install

All three deploy the same product. Differences are in operational complexity and infra footprint — not capability.

Recommended for production

Kubernetes (Helm)

EKS, AKS, GKE, OpenShift. Horizontal scaling, rolling upgrades, your ingress controller. Customer-managed Postgres or RDS.

Install:

helm repo add insightstudio https://charts.insightaistudio.com
helm repo update
helm install studio insightstudio/studio \
  --namespace insightstudio --create-namespace \
  --values your-values.yaml

A sample values.yaml is included with the chart — wire in your Postgres DSN, S3 bucket, Okta tenant, and a list of admin emails.

Air-gapped variant: we ship a signed offline image bundle. Customers with no outbound network registry pull it once, scan it, and load it into their internal registry.

For pilots / small teams

Single VM (docker-compose)

One Ubuntu / RHEL / Amazon Linux VM. Postgres + Studio API + UI + a default worker in one stack. Sub-30-minute install. Upgrade path to Kubernetes when ready.

Install:

curl -fsSL https://install.insightaistudio.com/studio.sh | bash
# OR download the offline bundle:
curl -fsSL https://insightaistudio.com/releases/insightstudio-latest.tar.gz -o studio.tgz
tar -xzf studio.tgz && cd insightstudio && ./install.sh

Edit .env for DB password, model provider keys, Okta tenant, then docker compose up -d.

Add workers to the fleet

Studio doesn't execute apps itself — your worker boxes do. Any machine with the InsightWorker CLI installed can join the fleet:

1. Install the CLI

curl -fsSL https://insightworker.com/install.sh | bash

2. Join the fleet

insightworker --worker \
  --studio https://studio.acme.com \
  --token $STUDIO_TOKEN

Worker boxes can be anything: engineers' laptops, dedicated EC2/Azure VMs, on-prem Linux servers, or k8s pods. They register with Studio, advertise their installed skills + credentials, poll for queued jobs, and run them locally. Token is provisioned in Studio's admin console.

Run as a service: standard systemd / launchd unit files in the docs.

Prerequisites

Required infrastructure

Database: Postgres 13+ or MySQL 8+ (customer-managed)
S3: a private bucket (your AWS account) for the app catalog
SSO: Okta, Azure AD, or Google Workspace (OIDC)
TLS: a cert chain your browsers trust (your CA, ACM, Let's Encrypt, etc.)
Model provider: at least one of Bedrock, Azure OpenAI, OpenAI, on-prem GPU (OpenAI-compatible endpoint)

Optional

SIEM: for audit log forwarding (Splunk, Datadog, Elastic, etc.)
Container registry: if your infra forbids pulling from public registries
Sealed-secrets / external-secrets controller if running on k8s
Reverse proxy: any ingress that can do TLS termination + path routing

Configure your LLM provider

Studio routes inference through whichever provider its workers know about. Configure one (or several — capability matching routes per-app). All standard provider envs work:

AWS Bedrock

USE_BEDROCK=true
AWS_REGION=us-east-1
BEDROCK_MODEL=us.anthropic.claude-sonnet-4-5-20250929-v1:0

Azure OpenAI

AZURE_OPENAI_ENDPOINT=https://my.openai.azure.com
AZURE_OPENAI_API_KEY=...
AZURE_OPENAI_DEPLOYMENT=my-gpt-deployment

Custom (on-prem GPU)

CUSTOM_LLM_BASE_URL=http://gpu-box.internal:8000/v1
CUSTOM_LLM_API_KEY=...   # if your endpoint needs it
CUSTOM_LLM_MODEL=meta-llama/Llama-3.1-70B-Instruct

OpenAI / Anthropic direct

OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-4o
# or:
ANTHROPIC_API_KEY=sk-ant-...
AGENT_MODEL=claude-sonnet-4-5

Need help with the install?

Tell us about your environment and we'll walk you through a 30-minute pilot install on a sandbox VM.

Talk to us