LangSmith BYOC Expands to Azure and Google Cloud for Agent Teams

The infrastructure landscape for agent deployment shifted this week as LangChain announced that its LangSmith Bring Your Own Cloud (BYOC) architecture is now available generally on AWS, and in beta on both Azure and Google Cloud, according to the LangChain Blog.

What happened

According to the LangChain Blog, LangSmith BYOC splits infrastructure into two distinct components: a control plane managed by LangChain and a data plane owned by the organization. The control plane handles authentication, organization and workspace configuration, billing metadata, hashed API keys, and static UI assets. Meanwhile, the data plane—which can span multiple accounts and regions—retains all sensitive assets, including traces, prompts, datasets, experiments, agent deployments, and workspace secrets.

The expansion brings specific cloud-native integrations to each provider. On Azure, BYOC runs the data plane inside a dedicated virtual network featuring a private AKS API endpoint, utilizing Istio and KEDA with Helm for ingress routing and event-driven scaling, with system telemetry routed to Azure Monitor. On Google Cloud, data planes can run in any region while managed from LangSmith's US control plane, utilizing namespace-scoped names to prevent infrastructure collisions. On AWS, the data plane operates within an existing VPC, allowing security teams to attach custom security groups for ingress and egress inspection.

Previously, organizations with strict compliance requirements had to deploy and maintain self-hosted LangSmith instances, requiring dedicated platform engineers to patch stateful services, configure databases, and manage schema migrations. With BYOC, LangChain provisions and operates the application inside the organization's virtual network while handling scaling, schema updates, and runtime dependencies. Organizations can migrate from self-hosted environments using an open-source migration tool for workspace resources, supplemented by internal LangChain tooling for trace migrations.

Why it matters for small AI engineering teams

For small team leads and founders running agent applications in production, infrastructure maintenance is a constant tax on velocity. Until now, choosing strict data sovereignty meant paying a heavy operational toll: engineering hours spent wrangling Helm charts, managing database migrations, and patching stateful services. BYOC solves this by splitting the burden. You keep sensitive agent telemetry, prompts, and evaluation datasets safely inside your own cloud perimeter, satisfying enterprise compliance requirements without needing a dedicated platform engineering team to keep the lights on.

This architecture also unlocks multi-cloud flexibility for growing businesses. If your core product stack runs on AWS while an enterprise client mandates Azure for their isolated business unit, your platform team can provision separate data planes tailored to each cloud's security model while maintaining a unified workspace strategy. By abstracting away the runtime dependencies and scaling mechanics, this model allows smaller teams to punch above their weight class and deploy enterprise-grade agent infrastructure with minimal overhead.

What to do this week

1. Audit your current agent infrastructure and data residency requirements to determine if your sensitive traces and evaluation datasets must reside within a specific cloud provider's virtual network.
2. Review the public Terraform modules and architecture documentation provided by LangChain on GitHub to evaluate the implementation complexity for your team's primary cloud environment (AWS, Azure, or Google Cloud).
3. Contact LangChain sales if your current self-hosted setup is taxing your engineering bandwidth and you are ready to plan a zero-downtime migration to the BYOC model.

Sources

• LangSmith BYOC is now available on AWS, Azure, and Google Cloud — LangChain Blog