Self-Hosted Onboarding Overview
Understand what a LangSmith Self-Hosted deployment involves and how to work with your Solution Architect through the installation.
What is LangSmith Self-Hosted?
LangSmith Self-Hosted runs entirely within your own infrastructure. You manage the full stack — Kubernetes cluster, databases, storage, networking, and the LangSmith application itself. This gives you complete data control and the ability to run LangSmith in environments that cannot send data to external services.
This is the most complex deployment model. Every self-hosted customer has a dedicated Solution Architect who guides them through the full installation and configuration.
What a Production Deployment Involves
Before your SA engagement begins, it helps to understand what goes into a production self-hosted deployment:
Infrastructure
LangSmith runs on Kubernetes. Supported distributions include EKS (AWS), AKS (Azure), GKE (GCP), OpenShift 4.14+, and other CNCF-conformant clusters.
A production deployment requires externally-managed services for each of these components — do not use the in-cluster defaults for production:
| Component | Options | Notes |
|---|---|---|
| PostgreSQL | RDS, Cloud SQL, Azure Database | Stores metadata, RBAC, configuration |
| Redis | ElastiCache, Memorystore, Azure Cache | Required version ≥ 5.0 |
| ClickHouse | LangChain Managed · ClickHouse Cloud · Self-managed (Altinity operator) | In-cluster is dev/POC only — see options below |
| Blob Storage | S3, GCS, Azure Blob | Strongly recommended — storing payloads in ClickHouse degrades cluster performance |
ClickHouse production options
- LangChain Managed ClickHouse — LangChain-operated ClickHouse service, zero ops overhead for your team. Simplest path if data residency allows it.
- ClickHouse Cloud — ClickHouse Inc.'s managed cloud service. Good option if you need a cloud-hosted ClickHouse that you control directly.
- Self-managed replicated cluster — Run your own ClickHouse outside the LangSmith namespace. The Altinity Kubernetes Operator is the standard way to deploy and manage a replicated ClickHouse cluster on Kubernetes. Your SA can help size and configure this appropriately for your environment.
Application
LangSmith is deployed via Helm chart. Your SA will help you configure your values file, connect to external services, and set up post-deployment requirements including:
- Ingress and custom domain
- TLS certificate management
- SSO (OAuth 2.0 / OIDC) with your Identity Provider
- Environment separation and upgrade procedures
Deployment Resources
Self-hosted deployments take one of two paths depending on whether you're bringing an existing cluster or provisioning from scratch. Your SA will help you confirm which path fits your environment before installation begins.
Path A — Installing into an Existing Cluster
Most enterprise customers fit this path. Your platform, networking, and security teams already own the underlying Kubernetes cluster and surrounding infrastructure, and LangSmith is being installed into that environment alongside other workloads. The public LangSmith self-hosted documentation is your source of truth — start with the general setup guide and then layer in the cloud-provider-specific guidance. Your Solution Architect will walk through these with you, help configure your Helm values, and answer environment-specific questions.
LangSmith Self-Hosted — Public Documentation
The canonical LangSmith self-hosted setup guide. Covers architecture, prerequisites, Helm chart configuration, external services wiring, ingress, TLS, SSO, and post-install steps.
AWS Self-Hosted
EKS-specific guidance — IAM Roles for Service Accounts (IRSA), RDS, ElastiCache, S3, and ALB ingress.
View provider docs ↗
Azure Self-Hosted
AKS-specific guidance — Workload Identity, Azure Database for PostgreSQL, Azure Cache for Redis, and Blob Storage.
View provider docs ↗
GCP Self-Hosted
GKE-specific guidance — Workload Identity, Cloud SQL, Memorystore for Redis, and Google Cloud Storage.
View provider docs ↗Path B — Greenfield: Provisioning a New Environment
If you're standing up a fresh environment end-to-end — cluster, databases, networking, and IAM — the PS Portal includes opinionated Terraform modules per cloud provider that provision the full stack alongside LangSmith. These are most useful for POCs, dev environments, or accounts where the SA team is taking the lead on infrastructure provisioning.
Azure AKS
Full GuideAKS with Azure CNI, Workload Identity, NGINX ingress, cert-manager TLS, private PostgreSQL, Redis, and Blob Storage.
GCP GKE
Full GuideGKE Standard/Autopilot with Workload Identity, Envoy Gateway ingress, Cloud SQL, Memorystore Redis, and GCS.
AWS EKS
Full GuideEKS with IRSA, ALB Ingress Controller, RDS PostgreSQL, ElastiCache Redis, and S3 via VPC endpoint.
OpenShift OCP
Coming SoonOpenShift on-prem or ROSA with Routes, cert-manager, PostgreSQL Operator, Redis Operator, and ODF.
Even on the existing-cluster path, your SA may reference these modules as a configuration template for Helm values and external service wiring.
What to Prepare Before Your Engagement
Help your SA hit the ground running by having the following ready:
- Cloud provider and Kubernetes distribution confirmed
- Kubernetes cluster provisioned or access confirmed (existing infra path) or cloud account access ready (greenfield path)
- LangSmith License Key received from your Account Executive
- Container registry accessible from your cluster
- Identity Provider details for SSO configuration (Okta, Azure Entra ID, etc.)
- Networking requirements understood (egress requirements, private vs. public endpoints)
- Team contacts identified — who owns infrastructure, who owns the application, who owns security/IAM
Next: Configure Your Organization
Once your SA has completed the deployment, use the Admin Readiness Review to work through all org configuration — SSO, SCIM, workspace design, roles, ABAC policy, API keys, data retention, Fleet setup, and more. Your SA will guide you through each topic.