Local Development

How to build data models on your own machine with a coding agent, the VS Code extension or the cred CLI, keep them in Git, and publish them by hand or from CI/CD

You can build data models in your own editor, with your own coding agent, in files you keep in Git. A Credible data model is Malloy (opens in a new tab) code in plain files, so it fits the workflow you already use for code: edit on your machine, review in pull requests, publish from your terminal or from CI.

To build without installing anything, build in the Credible App instead. Both paths produce the same packages, served the same way.

How Local Development Works

  1. Set up your project directory. Install the VS Code extension in VS Code or Cursor, or run cred init with the Cred CLI in a directory for any coding agent. Either one signs you in to an environment, gives the agent modeling MCP tools and Credible's modeling skills, and lets it use the environment's connections. Credentials stay in Credible; none are stored on your machine.
  2. Build the model in files. The agent finds your tables and drafts .malloy files in the directory. cred init also runs a local Malloy Publisher (opens in a new tab) that serves the model as you edit, querying your real databases through Credible.
  3. Version and review in Git. Commit the package and review changes in pull requests.
  4. Publish. Run cred publish from the package directory, or let the CI/CD template publish on every merge to main.

The Malloy modeling and analysis skills your agent gets are the same ones the in-app agent uses, and they are open source in Malloy Publisher, so the agent works the same way in your editor as in the Credible App.

Next Steps

On this page