Deployments

Deployments represent a snapshot of a project at a specific point in time. This snapshot includes all the sources for a project (uploaded assets), and the generated outputs, such as logs and tool definitions. The tools in use by a project’s MCP servers are always based on the most recent successful deployment, known as the active deployment.

A project’s deployment history can be accessed from the Deployments page:

Creating deployments

A deployment is created whenever a new source is added, or an existing one is updated or removed — uploading an OpenAPI document, pushing a function, or adding an external MCP server all trigger a deployment. Deployments are immutable: each change produces a new deployment that carries forward the unchanged sources from the previous one.

Once the deployment process is finished, tool definitions generated from it can be used in toolsets. If a deployment includes updates to existing sources, any dependent tool definitions, toolsets, and MCP servers are updated automatically.

Deployment lifecycle

Each deployment moves through a fixed set of statuses:

• Created — the deployment has been recorded but processing hasn’t started.
• Pending — sources are being processed into tool definitions.
• Completed — processing succeeded, and the deployment becomes the active deployment.
• Failed — processing failed, and the previous successful deployment remains active.

Two actions on the Deployments page manage failures and rollbacks:

• Retry — re-run the latest deployment after a failure.
• Redeploy — re-run an older completed deployment, effectively rolling back to it.

Troubleshooting

Information about each deployment can be accessed from the Deployments page of a project. Deployment logs, sources, and tool definitions can be viewed after clicking on a specific deployment. Deployment logs are especially useful for debugging issues with sources that caused a deployment to fail, or discovering why a tool was skipped when being created or updated from a source.