Your setup / Code Dev Home

Know what stays here and what connects.

A desktop workspace can read local files while sending selected context to a remote model. The word local describes only part of that arrangement. Check the application, model endpoint, and each connector separately when deciding what information belongs in a session.

Follow the information through the setup

Identify where inference runs, which content enters requests, and what the provider says about retention. A connector may add another service or process with its own permissions. Confirm the relevant configuration and use a small, non-sensitive task for the first connection. Keep credentials in the documented storage mechanism.

Choose model behavior deliberately

Unfiltered and uncensored AI coding describe expectations that vary by model, provider, and application. Compare the same legitimate development task across the setups you are considering. Check context accuracy, useful uncertainty, and compatibility with your workflow. These previews do not guarantee unrestricted behavior, unlimited capacity, or an independently audited privacy posture.

Connection inventory

Adapt to your project

Application: Version and supported connection format.
Model: Provider, endpoint, and exact identifier.
Data: Context that may leave the machine.
Credentials: Documented storage method.
Connectors: Enabled services and permitted actions.
First check: One non-sensitive read-only task.

Does a local window mean local inference?

No. Confirm the configured model endpoint and application behavior. A desktop interface can connect to remote infrastructure, and connectors can create additional data flows.

Shared reference preview: a connector list inside settings. A listed service does not establish availability in the final product. Reference desktop preview; final product screens may differ.