# Terminal interfaces

Choose between the cogs Control Room and the separate mind chat application.

Source: https://docs.minds.sh/docs/developers/tui



Minds has two terminal surfaces in this checkout. They have different binaries and responsibilities.

## cogs Control Room [#cogs-control-room]

```bash
cogs control-room
```

The Rust CLI's Control Room organizes Home, Instances, Graph/Memory, Inference, Agentic, and Settings views. It can use saved instance endpoints and the CLI connection configuration. Run a normal subcommand with `--no-tui` when you need predictable script output.

The current Home view combines real health checks with a synthetic throughput display. Storage and developer-toolkit sections include placeholders. Do not use those panels as a measurement source or assume they establish full production monitoring coverage.

Source: `minds-cli/src/control_room/mod.rs` and its `views/home.rs`.

## mind chat application [#mind-chat-application]

The separate Go project in `minds-tui` builds a `mind` command. Its entry point opens an interactive chat interface using the selected provider configuration and a local application database.

```bash
# From the minds-tui source directory
go build -o mind .
./mind --cwd /path/to/project
```

The checked-in module requires Go 1.25.0. The legacy upstream module name is still present in source imports; it is not the product name.

| Command or flag                          | Behavior                                                                                    |
| ---------------------------------------- | ------------------------------------------------------------------------------------------- |
| `mind`                                   | Start interactive chat                                                                      |
| `mind --cwd /path/to/project`            | Select the working directory                                                                |
| `mind --debug`                           | Enable debug logging                                                                        |
| `mind run 'Explain this module'`         | Run one non-interactive prompt                                                              |
| `mind run --quiet 'Explain this module'` | Hide the spinner                                                                            |
| `mind logs --tail 100`                   | Read a bounded number of application log lines                                              |
| `mind logs --follow`                     | Follow log output                                                                           |
| `mind schema`                            | Emit the configuration schema; registered as a hidden command                               |
| `mind --yolo`                            | Automatically accept permission requests; use only when you deliberately want that behavior |

The root help example mentions `--dev`, but the inspected root parser does not register that flag. Use declared flags rather than treating a help example as an implemented connection flow.

## Configure the chat application [#configure-the-chat-application]

The loader merges its global configuration with `mind.json` and `.mind.json` in the selected working directory. `mind schema` emits the current configuration schema, including providers, selected models, MCP servers, language servers, permission settings, and application options. Use a model identifier currently offered by your configured provider.

A local MCP demonstration can be configured in `mind.json`:

```json
{
  "mcp": {
    "minds-local-demo": {
      "type": "stdio",
      "command": "/absolute/path/to/akasha-mcp",
      "args": ["--transport", "stdio", "--namespace", "local-demo", "--agent-id", "mind"],
      "timeout": 30
    }
  }
}
```

This is the `mind` application's `mcp` format, not the `mcpServers` shape used by Cursor and Claude Desktop. Its local MCP backend remains process-local. Local application data defaults to `.mind`, and debug logs are written under its `logs` directory.

Provider and MCP configuration are loaded by the terminal application's configuration system. A running chat and local SQLite session do not by themselves prove access to your dedicated Mind. Verify the configured memory tool's endpoint, capability, and persistence using [Agent integrations](/docs/agents).

Source: `minds-tui/internal/cmd/{root,run,logs,schema}.go`, `minds-tui/internal/config`, and `minds-tui/go.mod`.