# Choose how to work

Use conversation, visible controls, a terminal, or code for the same Mind.

Source: https://docs.minds.sh/docs/getting-started/choose-an-interface



You do not have to choose one interface forever. Use conversation to express intent, the workspace to inspect details, and code when a workflow needs to run repeatedly.

| Interface     | Best for                                                            | Before you begin                                      |
| ------------- | ------------------------------------------------------------------- | ----------------------------------------------------- |
| Helm          | Asking for help and working through supported actions in chat       | Sign in and select the correct workspace              |
| Workspace UI  | Inspecting stored data, graph relationships, activity, and settings | Select a ready Mind                                   |
| CLI or TUI    | Exploring commands and repeatable terminal workflows                | Check the command's actual connection behavior        |
| HTTP API      | Precise integration with an explicit request and response           | Instance endpoint and authorized capability           |
| SDK           | Typed helpers in an application                                     | Match the SDK version, route, and authentication mode |
| Agent harness | Letting an assistant invoke memory tools                            | Compatible transport and a durable backend            |

## Keep the boundaries clear [#keep-the-boundaries-clear]

The dashboard is a view of your Mind, not the source of its memory. Closing a browser tab should not be how you manage the lifetime of a ready instance. Conversely, a local MCP process can have its own in-memory state; it does not become durable merely because it is launched from a Minds setup guide.

Read [agent transport compatibility](/docs/agents) and [SDK compatibility](/docs/developers) before relying on a tool across sessions.