Lolly for Operators

A defence-in-depth security & intelligence strategy - that just happens to be a creative production platform

The zero-trust organizational immune system that wraps around what you already do - so the routine creative work your teams need every day happens inside your perimeter instead of leaking out of it.

What's in it for you. You get to be the person who said yes to something both safe and popular. You close an exfiltration hole, gain capability, and delete a request queue in one move - the rare security win that makes you more liked, not less. No 3am call from legal because embargoed files or customer data found their way into a random web tool; fewer SaaS vendors, contracts, and audits on your plate; and a fully reproducible audit trail you can point to when someone asks. You sleep better, and brighten a few days doing it.

Lolly is no second-class creative tool: it puts production-quality output in everyone's hands, and the brand-guided creation experience is second to none. The reason it is safe to hand out widely is architectural: nothing uploads that you didn't put there, every result is reproducible, and every export can carry multiple layers of industry-leading cryptographic records. No matter how a document reached your desk, you can see its full provenance, whether it has been tampered with, and whether you can recreate it pixel-perfect.

Where it stands today. Lolly's security properties are strong by design, and its cryptography and file-parsing engines are going through SUSE's enterprise-grade infrastructure hardening. The seals, on-device signing, and encryption below are real and defensible now, and maturing toward independent certification - so where a contract calls for certified assurance, deploy them as defence-in-depth while that process completes.

The strategic advantage

The usual way routine creative work gets done is a liability surface: files emailed to external design contractors, brand assets uploaded to a dozen SaaS editors, customer data pasted into a stranger's web tool to "just make a quick graphic." Every one of those is data leaving your control.

Lolly inverts it. The work that drove those leaks - the quote card, the localized banner, the event badge, the redacted screenshot - now happens on a tool that runs on the employee's own device, against your brand, with no server in the loop. You didn't add a control on top of a risky workflow; you replaced the risky workflow with one that has no exfiltration path to begin with.

You govern the whole relay. A creative authors the rules and a developer scales them, but it's the operator who makes that lifecycle safe to run org-wide - the same tool that lets a rep self-serve on a plane is one you can gate through Git review, deploy through your MDM and verify cryptographically. See how the roles compound in The lifecycle of a campaign, and how you govern it in Adoption & Governance.

Delete the request queue while proliferating content.

One goal of Lolly is design-request deflection: routine requests that never need to reach a designer because the person who needed the asset made it themselves, correctly, in minutes. Every deflected ticket is both a productivity win and one fewer file changing hands.

Lolly is built to fit how your organisation actually operates - there's no single right way to deploy it:

The full deploy models and administration walkthrough live in Deployment and Configuration.

Anti-exfiltration utilities

A category of Lolly tools - the privacy utilities - exists specifically to keep files inside the perimeter.

Remove location and all hidden identifying information from documents and media files.

Anonymize, encode, format, and manipulate structured and unstructured text.

Shrink an oversized PDF on-device, so nobody reaches for a third-party "compress my PDF" website the moment a file is too big to email - which is exactly where data slips out the window.

All of these are on-device transforms: your file or data goes in, cleaned bytes come out, and there is no server to upload to. They are the deliberate opposite of the typical "upload your file to a stranger's website to clean it" tool that a well-meaning employee reaches for otherwise.

Strip Hidden Data: the file lands on the canvas and the badge states plainly that nothing is uploaded

Text Helper is the same bargain for text rather than files. It is the tabbed workbench an employee would otherwise go looking for on a stranger's site, and it declares no inputs at all because nothing it touches ever leaves the page.

Text Helper's workbench - a rail of operation tabs above a card stating that nothing you paste leaves your device

Compress PDF completes the set: the oversized attachment shrinks under a quality setting you choose, on the machine that already holds it.

Compress PDF - a quality level and a greyscale switch on the left, a drop zone for your own PDF on the right, and no upload anywhere

Determinism & reproducibility

Every tool input is expressible as a URL parameter, and the same inputs produce the same file. That has two operator consequences:

Prompt to Image is determinism at its plainest: the text is the whole input, the typeset image is the whole output, and the same text always sets the same way.

Prompt to Image - a block of prompt text typeset into a square image, with nothing in the result that was not in the input

Provenance & Content Credentials

The Verify drop zone accepts any file, from any source, and reads it without a network call

Exports can carry Content Credentials - a signed C2PA manifest bound to a hash of the file's bytes. Any later change to the file breaks the seal, so a C2PA-aware verifier detects alteration cryptographically, offline. The credential is tamper-evident: it flags tampering rather than preventing it, which is precisely what makes fully offline verification possible.

Interoperability notes. Lolly verifies its own credentials and many third-party ones offline today, including reading C2PA claim v2 manifests from other producers. One interop item remains in progress: WebM - which has no standardised C2PA mapping yet, so Lolly attaches the manifest as a Matroska part (third-party tools verify Lolly's MP4 out of the box; WebM follows once the standard settles).

Encryption & passwording

For files that must travel locked, everything happens on-device:

The lock card in the export panel: a password, and an explicit choice between the two tiers

Air-gap ready

Air-gap is a first-class deployment, not a special mode - Lolly runs with no network at render time out of the box. The web shell is an offline-first PWA (service worker); fonts and WASM are stored on-device; tool state is persisted locally through the host bridge, never localStorage. Any tool that reaches the network does so only through an allowlisted host.net capability it must declare in its manifest - a shell that can't (or won't) fulfil it stubs it out. Ship the shells to devices through your MDM, or serve one instance inside your network, and a fully air-gapped install renders, exports, encrypts, and verifies credentials with nothing to phone home to.

Good to know

A few things worth having straight before you roll it out:

Where to go next