The desktop command center — more screens, more control

What you get on Windows

The VS Code extension already runs the local Host and opens curated boards in an editor tab. The optional Windows installer adds the desktop command center — it expands that Host with screens and capabilities the in-editor surface does not cover . Same Host, same data, same license and trial — more UI and the wire CLI. Nothing in the extension requires it.

• Beyond the editor boards — full Live Traffic and Session Timeline (including compare), Sandbox, Privacy Audit, the full custom-dashboard catalog, and Working samples / History on pipeline stages.
• Full Settings — provider keys, licensing, retention, alerts, and desktop updates in one place.
• CLI included — wire --help talks to the same loopback /v1 as the extension.
• Auto-updates — the desktop app updates from the Halo Softworks release feed (Settings → Desktop → Updates).

Inside the command center

These are the desktop boards — the expanded Host UI. VS Code docks a curated subset beside your code; this is where the additional screens live. Sanitized telemetry only; prompts are never stored.

Pipeline — Overview

Pick a goal — observe only, protect secrets, spend less, prefer local models, coach prompts, or the recommended stack — then see Editor → Guard → Terran → Router → Compress → Advise → Observe → provider. Stages run in memory; mutation stages stay off until you choose. Observe records sanitized telemetry when licensed (included in the trial; $39 after).

Why it matters: Understand the command center at a glance before enabling any mutation stage.

Live Traffic — Overview

Real-time traffic mix, event rate, token flow, image-vs-text input, idle burn, latency distribution, provider mix, and model mix — no prompt text displayed or stored.

Why it matters: Answer “what model am I hitting?” and “how busy is this session?” in seconds.

Traffic & routing

Prebuilt dashboard with token flow, requests by project, estimated Router savings, and provider mix — visibility and spend on one board.

Why it matters: See what your agents are doing and what it costs without opening five different views.

Cost Estimation — Overview

Combined and projected cycle spend, plan allowance, Direct API vs plan estimates, and Tips for reducing costs — from observed tokens, not provider invoices.

Why it matters: Know where the cycle is headed before the invoice arrives — including Plan usage capture when you sync a provider balance.

Waste & idle burn

Idle-burn minutes, tokens, and spend for the cycle, spend velocity, and per-IDE idle-by-process detail over a 4h window — plus cost-by-project and estimated-cost charts.

Why it matters: Your IDE spends tokens while you get coffee — catch the drip before it empties a plan allowance.

Pipeline health

Cross-stage overview — Guard blocks, Router matches, Compress savings, and Terran assess-to-enrich — on one board.

Why it matters: Monitor every mutation stage without hopping between Pipeline tabs.

Compression savings

Avg tokens saved per chat, requests compressed, and token/dollar savings by provider — from sanitized telemetry only.

Why it matters: Quantify context-aware compression in dollars, not just “it should help.”

Privacy Audit

Built-in canary test injects unique markers through the live pipeline, then scans local storage — every location comes back clean.

Why it matters: Verify the no-content-storage model — do not just take our word for it.

Request detail

Per-request structural metadata — provider, model, tokens, latency, and status — never prompt or response bodies. Structure and Findings tabs add charts and Guard outcomes.

Why it matters: See exactly what each stage did on a request without exposing message content.

Sandbox

Multi-chat composer to exercise Guard, Router, Compress, and Terran in dry-run (default) or optional live mode — without configuring an IDE.

Why it matters: Try before you proxy — validate pipeline behavior safely without touching editor settings.

Aura Guard — Stats

Blocked requests, block rate, top category, and inspected volume from sanitized telemetry — never prompt text.

Why it matters: Prove Guard is catching secrets on the wire without reading what you typed.

Aura Router — Stats

Route matches, tokens by route, before/after provider and model mix, and delegated volume — including Copilot CLI and cheaper-model downshifts.

Why it matters: Quantify routing wins in tokens and mix shift, not just a toggle.

Aura Advise — Advice

Prompt quality meter (clarity / structure / discipline) plus live advice cards from your own traffic — rule keys and counts only, never prompt text.

Why it matters: Catch vague prompts, retry storms, and costly habits as they form — without storing what you typed.

Aura Compress — Multimodal

Right-size oversized inline images and optionally distill screenshot-like frames to text with local vision — remote image URLs are never fetched.

Why it matters: Vision turns stop burning tokens on full-resolution screenshots you never meant to re-send every turn.

Settings — Alerts

Daily budget caps and idle token-burn warnings — banners appear in the top bar when thresholds are crossed.

Why it matters: Set spend guardrails without sending usage data to the cloud.

Download the Windows installer

Installs the desktop command center the local Host, and the wire CLI. It also adds the VS Code extension to editors it detects. Prefer the extension alone? Use Get Aura Wire for the .zip / .vsix.