Sign in Request a demo
[ Security & trust ]

A tool that protects your data — and keeps it yours.

Tracehold watches what your team sends to AI, so the first question you'll ask is what we do with that data. The answer is simple: the check happens on the device, prompts never leave the browser by default, everything is encrypted, and anyone wanting to see a sensitive item has to ask — and it's logged. It's all built and hosted in the EU.

Prompts stay in the browser Everything encrypted Built & hosted in the EU
Talk to our team See how we protect your data
[ Trust pillars ]

Built to protect your data — and stay out of your way

Four commitments that hold no matter which plan you're on.

On the device
Where the check happens

We see what's about to be sent to AI before it ever leaves the machine.

Private by default
What leaves the browser

The prompt itself never leaves the tab — we only learn the type of finding, its severity, and when.

Encrypted
Everything we do store

Tamper-proof, and every customer's data is fully isolated from every other.

EU
Built and hosted in Europe

So the tool that protects your data doesn't become a new data-transfer problem.

[ On-device detection ]

We check on the device. Your words stay yours.

The whole point of Tracehold is to keep sensitive data from leaving — so it would make no sense for us to take it ourselves. Detection happens right on the device, before anything reaches an AI tool. When we do need to keep a record, it's encrypted, and seeing it takes permission and leaves a trail.

Caught at the source

In the browser, the prompt your team types never leaves the tab to be checked — we look at it right there. When a file is scanned for sensitive content, we read it, return what we found, and then discard it. The document is never kept.

Encrypted, and only revealed on request

On the desktop, Tracehold can hold onto the exact text that triggered an alert — so an admin can tell a real leak from a false alarm. It's encrypted the whole time. No one sees it casually: revealing it takes a specific permission and is logged, so you always know who looked, when, and at what.

Even our own staff can't peek

Our internal tools are built to be harmless: they can only read, never change anything, and they can't open the encrypted content. The protection holds even from the inside.

Said honestly: the rule is "private by default, revealed only when someone with permission asks — and it's logged." On the desktop we deliberately keep the exact wording behind an alert so a real leak can be told apart from a false alarm — and that copy is encrypted and sealed until an authorised admin reveals it. We'll always tell you exactly where a sensitive item can exist and how it's protected.
[ See · decide · protect · record ]

What actually travels — and what stays put

A single check on the device decides everything. The network only ever carries the result, never the content.

Allow

Clean prompts pass straight through, unmodified.

ana@acme.io allowed

Observe

Flagged content is logged for visibility, without interrupting anyone.

ana@acme.io logged

Redact

The sensitive part is hidden or stripped out — so the rest can still go through.

ana@acme.io [email] redacted

Block

The whole thing is blocked, and the person gets a clear heads-up.

IBAN blocked
Our number-one design rule: the prompt never leaves the browser by default.
[ Encryption & signing ]

Everything is encrypted and signed

Anything we store is locked down and tamper-proof. Connection details to your other systems are never shown back to you in the clear, and a misconfigured server simply refuses to start. Secure defaults aren't an option you have to remember.

Encrypted at rest

Your stored records and every connection to your other tools are encrypted, and sensitive credentials are masked in our screens and reports — never handed back in plain text.

Tamper-proof by design

The data we hold is signed, so any change would be obvious. And the system refuses to run on weak settings — a server that isn't configured safely won't even boot.

Encrypted in transit too

Traffic between the parts of Tracehold is always encrypted end to end, including the link to its own database — there's no path where your data travels unprotected.

In plain terms: think of it like a safe that locks itself. Sensitive items go in encrypted, the lock leaves a tamper-evident seal, and the safe won't open at all if the keys aren't set up properly. Even the people who built it can't read what's inside — they can only see that the safe exists and that the seal is intact.

No master back door. No “trust us.” The protections are built into how the product works, not bolted on as a promise.

[ Tenant isolation · deny by default ]

Every customer's data is fully isolated

Your information lives in its own walled-off space. One company can never see another's data — not by accident, not on purpose. The separation is enforced deep in the system, not left to a setting someone might forget.

Walled off at the core

The line between customers is drawn at the deepest level of the system. If a request ever tried to reach across that line, it simply gets nothing back — the default is “deny,” not “allow.”

Built for groups of teams

Manage several teams or business units? A parent can look across its own group, while changes stay locked to where they belong. Viewing a team never lets you alter it.

Defence in depth

We don't rely on a single gate. The app keeps each customer's data apart, and the system underneath enforces it again — so even a slip-up is contained, not exposed.

The short version: your data is yours and nobody else's, and that's guaranteed by how the platform is built — not by hoping every query behaves. See the platform architecture →
[ Access & signed updates ]

Only the right people get in — and only trusted code runs

Sign-in is protected against the attacks that actually matter, and every update to Tracehold is verified before it can reach your team. A security tool has to be at least as trustworthy as the thing it protects.

Sign in with your own identity provider

Use Microsoft Entra or Okta to log your people in, tied to your verified company domain. Sign-in is hardened, with brute-force protection and optional multi-factor — only the right people get in, scoped to the right team.

Read-only keys for your tools

If you connect Tracehold to your own systems, the access keys are read-only and tightly scoped — they can look, never change. They open the same private space as a normal sign-in, nothing wider.

Every update is signed and verified

The browser extension, the desktop agents and the detection rules are all signed before we'll send them out, and verified on the way in. The product won't accept anything unsigned, and an old version can't be slipped back in to reintroduce a known problem.

Works with the tools your team already uses

Tracehold covers the AI tools your team already uses — including the ones nobody approved — across the browser, the desktop and developer and terminal tools. It rolls out as a browser extension you push with your normal device management: no proxy, no network surgery.

Most teams are protected in 2–4 weeks, with honest, published pricing — Free, Insight (€40/user·mo) and Governance (€60/user·mo).

Live compliance mapping you can show the board. Beyond protecting your data, Tracehold maps AI risk to GDPR, NIS2, ISO 27001 and SOC 2 as a clear, always-current view — so you can see where you stand and prove it. See compliance in depth →
[ FAQ · trust questions ]

The questions buyers ask about trust

Do you store our prompts or our data?
By default, no. The check happens on the device, in the browser, and the prompt itself never leaves the tab. We only learn the essentials — what type of thing was caught, how severe it was, and when. When a file is scanned, we read it, return what we found, and discard it. On the desktop, Tracehold can hold the exact text behind an alert so an admin can tell a real leak from a false alarm — but it's encrypted and only revealed when someone with permission asks, and that's logged.
Could your own staff read what our team typed?
No. Anything sensitive we keep is encrypted, and the only way to see it is the audited reveal — which takes a specific permission and is logged. Our internal tools are deliberately limited: they can only read, never change anything, and they can't open the encrypted content. The protection holds even from the inside.
How do you keep one customer's data from another's?
Every customer's data is fully isolated, in its own walled-off space, enforced at the deepest level of the system. One company simply can't reach another's information — the default is “deny.” And we don't rely on a single gate: the app keeps things apart and the system underneath enforces it again, so even a mistake is contained rather than exposed.
How do you stop a bad update from reaching the agents?
Everything we ship — the browser extension, the desktop agents and the detection rules — is signed before it leaves us and verified before it's applied. The product won't accept anything unsigned, and an older version can't be slipped back in to reintroduce a known problem. The thing that updates a security tool has to be at least as trustworthy as the tool itself.
Can a false-positive check override a confirmed secret?
No. Once something is structurally confirmed — a real API key, a card number that passes the Luhn check, a private key — no secondary check can downgrade it. Critical and high-severity findings carry a hard action floor: the block or redaction happens regardless. Secondary checks can only rule out a false alarm on lower-severity, ambiguous findings; they can never wave through a confirmed leak.
Where does our data physically live?
Tracehold is built and hosted in the EU, with data residency in Europe — so the tool that protects your data doesn't become the next cross-border data-transfer headache. We publish our subprocessors and DPA up front, so you can see exactly which providers touch your data and where they're based.
How fast can we be up and running?
Weeks, not months. It's a browser extension you push out with your normal device management — no proxy, no network changes, no big rollout project. Most teams go from first call to protected in 2–4 weeks, and the pricing is published up front: Free, Insight (€40/user·mo) and Governance (€60/user·mo).

Send us your security questions

We'd rather answer the hard questions up front. Walk through how your data is protected with our team, then start in observe-only mode — and see exactly what Tracehold would catch, with zero disruption to your people.

Talk to our team See pricing