Reference

Fix a connection or deployment problem

Find the next useful check when your agent or app cannot connect.

An app with its address, activity, and resources in one place

The agent cannot see Shipped

  • Confirm that Node.js 24 and the connector dependencies are installed.
  • Check all three absolute paths in the configuration.
  • Use local stdio transport. The portal URL is not a remote MCP endpoint.
  • Confirm SHIPPED_API_TOKEN is available to the agent process, then restart it.
  • Check that the token has not expired or been revoked and belongs to the intended workspace.

The deployment needs an action

Read the returned reason. It may need deployment access, a compatible resource, more available credits, or a revised spending limit. Ask the agent to explain the exact change and cost before approving it.

A stale plan or quote needs a fresh review. Repeating the old approval does not update its terms.

The tool stopped waiting

A timeout in your coding agent does not prove the deployment failed. Ask it to check deployment_status using the operation ID. Open Deployments in the portal to see whether work is queued, running, succeeded, or failed.

The app returns a connection error

Check the active release and application logs first. Verify that your app listens on the configured port and that its health route returns the expected response.

For intermittent 502 errors, record the URL, time with time zone, and response status. Check server keep-alive settings against the load balancer’s idle timeout. An error by itself does not identify which layer closed the connection.

If an address change is in progress, inspect its status before trying another name. Avoid repeated deploys while the original operation is still running.

Include useful details

When asking for help, include the project ID, operation ID, error code, and time. Remove tokens, passwords, database connection strings, and customer data from logs or screenshots.