The short answer is yes: **enterprise-grade encryption, your data is never shared with third parties, agency clients get isolated environments for each brand, and you maintain full ownership of your data at all times.**

The rest of this article is the part that is more useful than a reassurance — **what you control**, and where to look to verify it yourself.

[Watch the security walkthrough](https://kb.trackiq.com/kb-assets/kbvideo-security-explainer-da9f287c.mp4)

*Prefer to watch? Seventy-two seconds on what you control, and where to verify it yourself.*

## **What TrackIQ can do with your Amazon account**

**It reads. It does not sell, list, or message customers.**

Each connection is a standard Amazon authorisation: you sign in to Amazon, approve access, and are returned to TrackIQ. **TrackIQ never receives your Amazon password** — that exchange happens on Amazon's side, and what comes back is a revocable access token.

![Each source is authorised separately, and shows when it last synced](/kb-assets/shot-conn-detail-c876cc.png)

Two properties of that arrangement are worth knowing:

**Access is separable.** Seller, Advertising, DSP and AMC are four independent authorisations. You can grant advertising access without granting DSP.

**Access expires by design.** Amazon issues time-limited tokens rather than a permanent key, and revoking access in Seller Central or the Advertising console cuts it immediately. You are never in a position where you cannot withdraw it. See [**Fixing a Broken or Expired Amazon Connection**](https://kb.trackiq.com/hc/knowledge-base/articles/fixing-a-broken-or-expired-amazon-connection).

## **Who inside your company can see what**

Access is enforced per role, and the enforcement is genuine rather than cosmetic.

**Sections you do not have access to are removed from the sidebar entirely**, not greyed out. Navigating directly to the URL returns a **403 Access Denied** page rather than the content — the restriction is on the server, not in the menu.

Billing is a good example: a user without billing permission cannot reach Billing & Usage by any route, on any plan.

Users and roles are managed under **Users & Access**, available on Starter and above.

## **Multiple brands are separated**

Every reporting screen is scoped to the brand selected in the top bar, and **agency clients get isolated environments for each brand**.

This is the property that matters most if you manage other people's businesses: one client's numbers do not appear in another client's view, and access can be granted per brand.

## **AI access is the part worth checking**

This is where a security question actually earns its answer, because it is the newest surface and the least familiar.

**Aria** works inside TrackIQ, on the brand you have selected, under your own permissions. It cannot see what you cannot see.

**MCP** — connecting your own AI assistant, like Claude Desktop — has three properties worth stating explicitly:

**It is read-only.** Nothing an assistant does through MCP can change a bid, pause a campaign, or alter your account.

**It is scoped to one brand**, chosen on the consent screen when you connect. A connector approved for one brand cannot read another. Connecting a second brand is a separate, deliberate act.

**Every request is logged.** Settings → MCP & API Keys → **Call Logs** records the question that was asked, how many tool calls it took, and when.

![Call Logs is a per-question audit trail of AI access](/kb-assets/shot-mcplog2-header-fa1a52.png)

That last one is the practical answer to *“which AI tools have been talking to our data?”* — you can read it rather than trust it.

## **API keys are the one thing you must manage**

Everything above is enforced for you. API keys are the exception, because they are standing credentials that live wherever you put them.

**Issue one key per thing that uses it**, never one shared key — revoking a shared key breaks everything at once, while a named key breaks only what you meant to break.

**Never paste a key into a chat, a ticket, or a shared document.** Use environment variables or a secret store.

**Revoke rather than forget.** Each brand's panel shows active and revoked counts, which is your audit trail.

**If you only use Claude Desktop, issue no keys at all.** It self-registers through OAuth and needs none, which is the safest state available.

See [**Installing TrackIQ MCP in Claude Desktop and Cursor**](https://kb.trackiq.com/hc/knowledge-base/articles/installing-trackiq-mcp-in-claude-desktop-and-cursor).

## **A short checklist**

Worth running once a quarter, and it takes about ten minutes:

1. **Settings → Users & Access** — does anyone here no longer work on this account?
2. **MCP & API Keys** — is every active key still in use? Revoke the rest.
3. **Call Logs** — does the AI access match tools you recognise?
4. **Connections** — is anything connected that you no longer need?

The first and last catch the most common real-world exposure by a wide margin: **access that outlived the reason it was granted.**

## **Related**

- [**The MCP Web Console**](https://kb.trackiq.com/hc/knowledge-base/articles/mcp-web-console) — issuing and scoping AI access.
- [**MCP Call Logs: Seeing What Your AI Tools Asked**](https://kb.trackiq.com/hc/knowledge-base/articles/mcp-call-logs) — the audit trail for AI queries.
- [**Managing Account Settings**](https://kb.trackiq.com/hc/knowledge-base/articles/managing-account-settings) — where the account controls live.