Projects and channels

A project owns source code and one or more channels. A channel is a deployment environment, such as production or a preview, with its own live address.

Understand the scope

Every project starts with a prod channel. Each additional channel uses another slot in the owner's account-wide channel allowance. A preview is not a separate billing plan. CPU, memory, database usage, and file storage are account-scoped resources; adding projects does not multiply that budget.

Code-running channels have their own Postgres database. Static channels do not. Use the platform-provided environment values instead of deriving credentials or hostnames yourself. Platform sites use xhostd.com; tenant app addresses use xhostd.app.

Create and inspect a project

Connect an agent using the setup guide. Ask it to create a project with the template your workload needs. Choose a working recipe before writing the build and launch scripts.

Open Projects in the console. A project's Overview separates the currently serving channels from the latest deployment attempt. A failed attempt does not by itself mean the previous version stopped.

Use a preview channel

Ask your agent to create a separate preview channel and deploy a revision to it. Review the returned URL before deploying that revision to production. Verify the channel's environment values and data setup; do not assume a preview holds a current copy of production data.

See the channel API for the exact creation and deletion contracts. A new channel and a new code deployment are different operations.

Work on a shared project

Check the owner shown next to the project name. The API and MCP tools accept owner-qualified names where documented; the console keeps that owner context in links. Membership is not ownership. A member must not interpret an unavailable owner-wide resource summary as zero usage.

Manage invitations and roles in the project's Access section. See Access and security.