Requirements and sizing
Kindo runs on any conformant Kubernetes cluster, with PostgreSQL, Redis, object storage, and the other services it connects to managed or self-hosted. Infrastructure covers what each one needs. Every item is required unless it’s marked recommended or optional.
Choose a size
Section titled “Choose a size”| Size | Active users | Agent runs per day |
|---|---|---|
| Medium | 50 to 200 | About 1,500 |
| Large | 200 to 1,000 | About 5,000 |
Cluster capacity
Section titled “Cluster capacity”| Resource | Medium | Large |
|---|---|---|
| Cluster nodes (total) | 12 nodes: 96 vCPU, 192 GB RAM | 20 nodes: 320 vCPU, 640 GB RAM |
| Application nodes | 6 × 8 vCPU, 16 GB RAM | 10 × 16 vCPU, 32 GB RAM |
| Sandbox nodes | 6 × 8 vCPU, 16 GB RAM | 10 × 16 vCPU, 32 GB RAM |
| Block storage | 1.5 TiB | 3 TiB |
Required service sizes
Section titled “Required service sizes”| Resource | Medium | Large |
|---|---|---|
| PostgreSQL: main | 2 vCPU, 8 GB RAM | 4 vCPU, 16 GB RAM |
| PostgreSQL: auxiliary (recommended) | 2 vCPU, 8 GB RAM | 2 vCPU, 8 GB RAM |
| Redis | 3 GB RAM | 6 GB RAM |
| RabbitMQ (your own) | 2 vCPU, 8 GB RAM | 4 vCPU, 16 GB RAM |
Sizing notes
Section titled “Sizing notes”- Node counts are the maximum each node group needs. With a cluster autoscaler, set them as the node group maximums.
- Sandbox nodes run agents’ tools and scripts. Add sandbox nodes if you expect more agent runs.
- Block storage comes from your StorageClass and grows as people keep persistent agent workspaces. Volumes can grow where the StorageClass supports it, and can’t shrink.
- An auxiliary PostgreSQL server is recommended for production. Without one, its databases go on the main server, so add its size to the main server’s.
- RabbitMQ run by Kindo uses 4 GiB of memory and up to 1 CPU on the application nodes, within the sizes above.
- Required services you run in the Kindo cluster add to the cluster capacity. Add their resources, for every replica, to the totals.
Start from these sizes, monitor usage, and adjust as it grows. See Scaling and resources.
Network
Section titled “Network”Routing, DNS, and TLS
Section titled “Routing, DNS, and TLS”Get these ready before you install, so people can reach Kindo as soon as installing finishes.
-
Choose how traffic reaches Kindo. With Kindo-managed ingress, installing creates an Ingress for every host through your ingress controller. Choose it unless your load balancer restricts traffic to an allowlist. With your own routing, you create a route for every host once Kindo is installed. Ingress, DNS, and TLS compares the two.
-
Pick a base domain you control. Kindo serves these hosts under it:
Subdomain Serves app.The Kindo web app superadmin.The superadmin dashboard api.The Kindo API and inbound webhooks integrations-api.The integrations API integrations-connect.Sign-in pages for connecting integrations hatchet.Hatchet’s API and dashboard unleash.The Unleash feature flag service unleash-edge.Unleash flag delivery to the web app flagsmith.The Flagsmith feature flag service flagsmith-edge.Flagsmith flag delivery to the web app hyperdx.The HyperDX dashboard qdrant.Vector search, when Kindo runs Qdrant storage.File downloads, with Azure Blob Storage -
Create the DNS records. Point one wildcard record for the base domain, or one record per host, at the load balancer that serves your ingress controller or your own routes. Use an A record for an IP address and a CNAME record for a hostname. Your ingress controller’s Service shows its load balancer’s address:
Terminal window kubectl get service -n <ingress-controller-namespace>With Azure Application Gateway, point the records at the gateway’s frontend IP addresses, as its example describes.
If your controller creates its load balancer only when Kindo’s Ingress objects appear, as the AWS Load Balancer Controller does, have external-dns in the cluster publish the records, or create them as soon as installing finishes.
-
Get TLS ready. Have one of these for every host:
- cert-manager installed, with a cluster issuer that’s Ready. With Kindo-managed ingress, installing stops if the issuer isn’t Ready.
- A certificate and its private key, such as a wildcard certificate for the base domain.
- With the AWS Load Balancer Controller, an AWS Certificate Manager certificate.
- A certificate already on your load balancer or gateway.
Inbound
Section titled “Inbound”All inbound Kindo traffic is TCP. Open only the ingress entry points to the internet. If Kindo is internal-only, open them only to your corporate network.
| Port | Purpose |
|---|---|
| 443 | HTTPS: all user traffic |
| 80 | HTTP redirect to 443 (optional, recommended) |
Outbound
Section titled “Outbound”If the cluster can’t reach these destinations, installing fails, and so does Kindo itself. Confirm them with your network team before you provision.
| Port | Destination | Purpose |
|---|---|---|
| 443 | registry.kindo.ai and other container registries, AI model providers, external object storage | Image pulls, model inference, file storage |
| 5432 | PostgreSQL | Database traffic |
| 6379 | Redis | Cache and streaming |
| 5672 | RabbitMQ, if you run your own | Message queue |
| 25, 587, 465 | Mail relay |
If you forward audit events, also allow traffic to your syslog receiver: UDP or TCP 514, or TCP 6514 for TLS.
Access and security
Section titled “Access and security”Access controls in your environment are the most common reason installing is blocked. Arrange these with your security team early.
- Identity that installs Kindo. It needs permission to create and update namespaces, service accounts, role bindings, CRDs, and other cluster-scoped resources.
- Organization guardrails. Review policies that could block installing Kindo or the resources it creates: service control policies (AWS), organization policies (GCP), or management group policies (Azure).
- Workload identity. Services that call your cloud provider, such as LiteLLM calling AWS Bedrock, use workload identity. That’s IAM roles for service accounts (AWS), Workload Identity (GCP), or Microsoft Entra Workload ID (Azure). On AKS, enable the cluster’s OIDC issuer and Workload Identity, which the S3 gateway needs too; see Azure.
- Pod Security. If your cluster enforces Pod Security, allow the log collector and the sandbox init container the settings in the table below.
- Access to the config. The config contains sensitive credentials. Limit who can read the config Secret, and who can open the config to edit it.
- Container images. Kindo pulls images from the Kindo registry with the registry login Kindo provides. Allow the registry, or your internal mirror if air-gapped, in your registry allowlist and admission controllers. If you require scanned images, ask Kindo for software bills of materials or a mirror-and-scan workflow.
- Secret store. Choose where Kindo keeps its credentials: Kubernetes Secrets (the default), Vault, AWS Secrets Manager, or Azure Key Vault. See Secret store.
- Encryption. For production, encrypt PostgreSQL, Redis, object storage, and Kubernetes Secrets at rest, and PostgreSQL connections in transit.
| Item | Setting |
|---|---|
| Pod Security levels | baseline, restricted |
| Log collector | Runs as root, with host log paths |
| Sandbox init container | Runs as root, with the NET_ADMIN, NET_RAW, CHOWN and FOWNER capabilities |
| Config Secret | kindo-secrets-config in kindo-system |
| Registry host | registry.kindo.ai |
Production recommendations
Section titled “Production recommendations”- On-demand nodes. Use on-demand nodes. Spot nodes can be reclaimed at any time, which interrupts running agents and restarts stateful services.
- High availability. Run PostgreSQL with streaming replication or as a managed service with automatic failover. Run RabbitMQ as a cluster of 3 or more nodes with quorum queues, or as a managed service.
- Backups. Take daily automated PostgreSQL backups with 7 or more days of retention. Back up the config as described in Back up and restore.
- Disk alerts. Alert on disk usage on every PostgreSQL server, especially Hatchet’s, and on the cluster’s persistent volumes, before a volume fills.
Prerequisites checklist
Section titled “Prerequisites checklist”Each item names the team that typically owns it, so you can split the list across teams.
| Owner | Item | |
|---|---|---|
| ☐ | Platform | Size chosen (Choose a size) |
| ☐ | Cloud Ops | Cluster provisioned (Kubernetes cluster) |
| ☐ | DBA / Cloud Ops | PostgreSQL and Redis provisioned, and RabbitMQ if you run your own (Infrastructure) |
| ☐ | Cloud Ops | Object storage bucket created; with Azure Blob Storage, also the managed identity, its Storage Blob Data Contributor role and its federated credential (Object storage) |
| ☐ | Platform | Mail relay details in hand (Email) |
| ☐ | Security / AppDev | At least one model endpoint ready (AI models) |
| ☐ | Networking | Routing chosen, base domain picked, DNS records ready, TLS ready (Routing, DNS, and TLS) |
| ☐ | Networking / Security | Inbound and outbound firewall rules approved (Inbound, Outbound) |
| ☐ | Cloud Ops / Security | Identity that installs Kindo, guardrails, workload identity, registry, secret store, and encryption arranged (Access and security) |
| ☐ | Platform / DBA | Production recommendations planned (Production recommendations) |
| ☐ | Platform | Workstation ready (Prepare your workstation) |
