LangSmith BYOC is now available on AWS, Azure and Google Cloud

Separating control from data

According to a November 3, 2026 announcement from LangChain, LangSmith BYOC is generally available on AWS and in beta on Azure and Google Cloud. The setup lets companies choose the cloud and region where their agents run and where telemetry data is stored.

The system divides LangSmith into two parts. The control plane is operated by LangChain and includes authentication, organization and workspace configuration, billing metadata, hashed API keys and the static user interface. The data plane runs in the user’s own cloud environment, where tracing data, prompts, datasets, experiments, agent deployments and workspace secrets remain.

An organization can maintain multiple data planes in separate accounts and regions. They can be separated by region, environment or business unit. For example, production and test environments can run in separate AWS accounts, while a business unit standardized on Azure can use its own BYOC deployment.

LangChain operates the data plane in the cloud

BYOC is designed for teams that work under strict compliance requirements. Previously, these companies could maintain self-hosted LangSmith environments, but doing so required platform engineers to repair stateful services, configure databases and manage schema migrations.

According to LangChain, with BYOC the infrastructure remains within the company’s own virtual network while the provider manages scaling, schema changes and runtime dependencies. Teams therefore do not need to schedule maintenance windows for every update or manage custom Helm deployments. According to the announcement, new LangSmith features reach BYOC data planes shortly after they become available in LangSmith Cloud.

Based on its experience with AWS, LangChain highlights three main benefits: continuous updates, automatic scaling and faster access to the latest features. Services scale according to demand. The company said that many teams have chosen this operating model since the AWS BYOC launch in August.

Different approaches across the three clouds

In all three supported clouds, BYOC relies on the respective provider’s networking, identity management and security building blocks. Each data plane also runs Managed Deep Agents endpoints.

• Azure: the data plane runs in a dedicated virtual network with a private AKS API endpoint. LangChain deploys Istio and KEDA using Helm, while system telemetry is sent to Azure Monitor. Support for Azure LangSmith Sandboxes will arrive later.
• Google Cloud: at launch, the data plane can run in any region, while it is managed from LangSmith’s United States-based control plane. Resources are given namespace-scoped names so they do not conflict with existing infrastructure.
• AWS: BYOC runs in the company’s existing VPC, with the database placed in the selected subnets. Security teams can attach their own security groups to application roles.

It is also possible to migrate from self-hosted LangSmith. The open-source migration tool transfers users and roles, datasets, experiments and prompts, as well as settings for annotation queues, automation rules, dashboards and Fleet resources, among other items. The source system remains operational during the process. LangChain handles the migration of traces, runs and feedback in a separate step, which is currently available on AWS.

The Terraform modules behind BYOC and the migration tool are open source. The LLM Gateway running on the data plane calls models from 11 providers by default, while any OpenAI- or Anthropic-compatible endpoint can also be used. To request a BYOC data plane, customers must contact LangChain’s sales team.