Getting Started6 min read

The Interface

Chat panel, task header, mode chip, and everything on screen

Nexgile Code’s surface area is deliberately small: one chat panel, a handful of toolbar buttons, and a few integration points in the editor and terminal. This page names every piece so the rest of the documentation can refer to them precisely.

The chat panel

The panel (opened from the Nexgile Code icon in the activity bar) is a conversation view with an input area at the bottom. The input area packs in more than it first appears:

The input box — placeholder text reads “Type your task here…” with the hint “@ to add context, / for commands.” Multi-line input is fine; Enter sends.
@ mentions — type @ to open a picker over Problems, Terminal, URL, folders, files, and Git commits. Mentioning pulls that content into the task’s context directly, instead of hoping the agent finds it. Full syntax in Context Mentions.
/ commands — type / to run slash commands: the built-in /init plus any project or global commands you’ve defined. The same popup also lets you jump between modes.
The mode selector — a chip showing the current mode (each mode carries an emoji badge in the list, e.g. 💻 Code, 🏗️ Architect). Click it for a searchable list of all 31 built-in and specialist modes, described in Using Modes and Specialist Modes.
The API configuration selector — next to the mode selector; switches the active provider profile mid-conversation. Profiles can be pinned to the top of this list, and a lock option forces one profile across all modes in the workspace.
The auto-approve control — shows the current approval posture (“Auto-approve off,” or a count of enabled categories) and opens the toggle list. Details in Auto-Approving Actions.
Enhance Prompt — a wand button that rewrites your draft request with added clarity and context before sending.
Images — when the selected model supports image input, an add-images button appears, and you can paste an image from the clipboard or hold Shift to drag files/images straight into the box. Screenshots of broken UI make excellent bug reports to the agent.
Message queueing — if you type while a task is still running, your message is queued and sent when the current work completes, so you can stack instructions without interrupting.

The panel’s title bar carries the task-level controls:

Button	What it does
New Task (+)	Starts a fresh conversation with fresh context.
Settings (gear)	Opens the settings view — Providers, Modes, Auto-Approve, Checkpoints, and the rest. See Settings & Feature Flags.
Task History (under …)	Every past task, fuzzy-searchable, sortable by Newest, Oldest, Most Expensive, Most Tokens, or Most Relevant. Reopen, export, copy the prompt, or delete tasks from here.
Open in Editor (under …)	Moves the chat into a full editor tab — same conversation, more room. Also available via the Command Palette as Nexgile Code: Open In New Tab.

The task header

At the top of an active conversation sits a collapsible Task header. Collapsed, it shows the task text and headline numbers; expanded, it breaks out:

Tokens — input/output token counts, updating live as the task runs.
Cache — prompt-cache reads and writes, where the provider supports caching.
API Cost — the running USD cost of this task. This is the number to glance at before approving “one more retry.”
Context Length — how much of the model’s context window is in use, with a progress bar. When it climbs, context management kicks in; you can also trigger Intelligently condense context manually from the header.

The header also carries per-task actions, including Export task history (save the conversation to a file) and task deletion.

The todo list display

On multi-step tasks the agent maintains a todo list, shown in chat as a progress line — “2 of 4 to-dos done” — that expands into the full checklist with completed and in-progress items. It updates as the agent works and is your quickest read on where a long task stands. How the agent uses it: Subtasks & Todo Lists.

Editor integrations

Nexgile Code is also reachable from the code you’re looking at:

Right-click submenu. Select code, right-click, and open the Nexgile Code submenu: Add To Context (attach the selection to your current chat), Explain Code, and Improve Code (each starts a focused task about the selection).
Lightbulb Code Actions. Nexgile Code registers quick fixes, including Fix Code on diagnostics — click the lightbulb (or press Ctrl+.) on a problem line. These are gated by the VS Code setting nexgile-code.enableCodeActions (on by default).

Terminal integrations

Right-click inside the integrated terminal for the Nexgile Code submenu there:

Add Terminal Content to Context — attach recent terminal output to your chat.
Fix This Command — hand a failing command to the agent to diagnose and correct.
Explain This Command — get a plain-language explanation of what a command does.

Keyboard shortcuts

Shortcut (Windows/Linux)	macOS	Action
`Ctrl+K Ctrl+A`	`Cmd+K Cmd+A`	Add To Context — sends the current editor selection to the chat. Requires an active selection.
`Ctrl+Alt+A`	`Cmd+Alt+A`	Toggle Auto-Approve — flips auto-approval on/off without opening settings.

Both are standard keybindings, rebindable in the Keyboard Shortcuts editor.

Command Palette

Everything the toolbar does is also reachable from the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) — type “Nexgile” to filter. Useful when your hands are on the keyboard or the panel isn’t open:

Command	What it does
New Task	Opens the panel and starts a fresh task with the input focused
Task History	Jumps straight to the history view
Open in Editor / Open In New Tab	Moves the panel into a full editor tab — far more comfortable for reading long diffs
Settings	Opens the settings view
Focus Input Field	Puts the cursor in the chat box from anywhere in the editor
Accept Input/Suggestion	Confirms the current input or highlighted suggestion without reaching for the mouse
Toggle Auto-Approve	Same as `Ctrl+Alt+A`
Set Custom Storage Path	Relocates task history and settings storage
Import Settings	Loads a previously exported configuration file

Assign a keybinding to any of them in the Keyboard Shortcuts editor if you use one often.

Agent Workflow extras

If you’ve enabled features under Settings → Agent Workflow, additional surfaces appear:

Task Dashboard — productivity metrics on the home screen: task count, cost, tokens, and a daily activity chart. Shows up once you have three or more tasks in history.
Pipeline View — a Plan → Code → Review → Test strip above the conversation, with each stage marked pending, active, done, or skipped. The stages are read off what the task does (reasoning, first edit, file reads, command runs), so it tracks ordinary tasks too — not just Workflow Mode runs.
Task Templates — pre-built workflow templates (Feature, Bug Fix, Refactor, Security Audit) that drop a structured prompt into the input box for you to edit before sending.

These are UI feature flags — some on by default, some off. The full list and defaults are in Settings & Feature Flags.

UI languages

A language preference in Settings tells the agent to think and respond in your language — 18 options are available: English, Catalan, German, Spanish, French, Hindi, Indonesian, Italian, Japanese, Korean, Dutch, Polish, Brazilian Portuguese, Russian, Turkish, Vietnamese, Simplified Chinese, and Traditional Chinese. You can also simply write to the agent in your language and it follows your lead.