Skip to main content

Choose a deployment model

Glean places no constraints on where a custom connector runs — there's no required runtime or hosting pattern. You can use Glean's managed, containerized option, or run your own job or service and push data through the Indexing API. Choose where it runs before you build, because it determines how secrets, scheduling, and networking are handled.

There are two deployment models:

Glean-hosted

The connector is packaged as a Docker image and configured in your Glean deployment. Glean manages secrets, scheduling, and runtime for you.

Self-hosted (GCP or AWS)

The same container runs inside your own cloud account. You manage secrets, scheduling, and networking to Glean's indexing endpoints.

Compare the models

DimensionGlean-hostedSelf-hosted (GCP or AWS)
RuntimePackaged as a Docker image and configured in your Glean deployment.The same container runs inside your own cloud account.
SecretsManaged in the Glean cloud project.Stored in your own secret manager.
SchedulingCrawls run automatically, scheduled like a native connector.You handle scheduling and orchestration in your environment.
NetworkingManaged within the Glean deployment.Your environment must allow pushes to Glean's indexing endpoints.
Best forTeams who want Glean to manage runtime and scheduling.Teams with strict data-residency or network controls.

In both models, setup is driven by environment variables, and the connector is managed like any other connector once deployed.

How each model works

On a Glean instance, the connector is packaged as a Docker image and configured in your deployment:

  • Secrets are managed in the Glean cloud project.
  • The connector is deployed and scheduled like a native connector, so crawls run automatically.
  • Glean manages the runtime — you don't operate any infrastructure.

This is the simplest option when you don't have a specific requirement to keep the connector inside your own cloud.

Which should you choose?

Pick Glean-hosted if you want Glean to manage runtime and scheduling and you have no requirement to keep the connector in your own cloud. Pick self-hosted if data-residency or network controls require the connector to run inside your GCP or AWS account.

Next steps