Tools

Profiles & programs

Copies or merges the contents of one Windows user profile into another — documents, desktop, pictures, browser data and the rest — then catalogues the programs that were installed on the source machine so they can be reinstalled deliberately rather than from memory. Profiles on a disk pulled out of another PC are supported, so a dead machine’s drive can be read directly.

Drivers & unknown devices

Identifies the yellow triangles in Device Manager from their raw hardware IDs and asks Windows Update which newer drivers are on offer for the machine. Scanning, decoding and reporting are entirely read-only.

Why it is safe to run

• The source profile is never modified. The engine only reads it — there is no move or delete path.
• Nothing runs without a preview. A live migration stays disabled until a dry run has completed for the exact settings you are about to use.
• Nothing is overwritten by default. Everything lands in a new sub-folder, so collisions cannot happen. When you deliberately merge in place, replaced files are backed up first.
• Every run can be undone. Each run writes a full log and a CSV manifest of every action, and the undo replays that manifest backwards.
• Dangerous configurations are refused — same source and destination, identical paths, or nested profiles are blocked outright.
• Copies are verified by size, with optional SHA-256 deep verification.

Scan the way you work

Choose flatbed or feeder, colour mode, resolution, page size and output format. Multi-page scans appear as thumbnails that can be reordered, rotated or removed before saving as one PDF or separate image files.

Private by design

• No cloud account, analytics or telemetry. Scans stay on your computer unless you choose to email them.
• No installer. Unzip the download and open the application.
• Existing files are never overwritten. A safe numbered name is chosen automatically.
• Saved defaults speed up repeat jobs. Folder, format, quality, source and naming pattern are remembered.

Pick once, install everywhere

Tick the applications you want and press install. PackPilot works through them one at a time, showing what it is doing and what happened to each one. Click a category heading to select or clear the whole group, and search by name, package identifier or licence.

The same selection can be saved as a profile, exported as a PowerShell script, a batch file or a winget import file on Windows, or as a shell script or Brewfile on macOS — so it can be driven from an RMM, Intune, a GPO startup script or a task sequence. On Windows it can also build a single deployment executable that carries your selection inside it.

Only from official sources

• No bundled installers and no adware wrappers. Windows installs go through the Windows Package Manager (winget), falling back to Chocolatey and then to the vendor’s own download. macOS installs go through Homebrew.
• Every package identifier is checked against the vendor’s own index before each release. The Windows list is validated against the winget community source index that winget itself downloads, and the macOS list against the official Homebrew cask and formula API. A name that does not resolve fails the build rather than the technician’s machine.
• Every application says how it is free. A badge marks each one as open source, freeware, free for personal use, or a free tier of a commercial product, so you know what you are putting on a client’s computer.
• One catalogue, both editions. The Windows and macOS builds read the same list, and the build fails if the copy inside either one has drifted from it.

What is in the catalogue

124 applications in 13 categories — browsers, messaging, media, runtimes, imaging, documents, security, cloud storage, file sharing, compression, utilities, remote access and developer tools. 110 can be installed on Windows, 90 on macOS, and 76 on both. You can add anything else by pasting its winget identifier or Homebrew token, or replace the list entirely with your own catalog.json.

Checking what you downloaded

About inside the application reports the version, build date, publisher, signature state and the SHA-256 of the running executable, so it can be compared with the checksum published above without a network connection.

Your originals are never touched

Everything you do is stored as a recipe beside the photograph, not baked into it. The file that came off your camera stays byte-for-byte as it was, and you can go back to it, or to any earlier point, at any time. Delete Lumen and every photograph you own is exactly where you left it.

Reads the raw files your camera makes

Lumen decodes raw through the same imaging engine macOS itself uses, so it opens Sony, Canon, Nikon, Fujifilm, Olympus and OM, Panasonic, Pentax, Leica, Hasselblad, Phase One, Sigma, GoPro and DNG files, alongside JPEG, HEIF, PNG and TIFF. Exposure recovery and highlight recovery happen while the sensor data still has room in it, which is detail no slider can get back later.

Fixes what the lens did

Lumen uses the camera maker’s own correction when your file carries one, matches a profile from its own database when it does not, and gives you manual control over distortion, corner darkening, colour fringing and perspective when nothing has ever measured your glass. It tells you plainly which of those three it is doing, rather than showing a tick box that quietly does nothing — and if it cannot identify your lens with confidence it says so instead of guessing, because the wrong correction is worse than none.

All the usual corrections, where you expect them

• Light. White balance, exposure, contrast, highlights, shadows, whites and blacks, with tone curves for each channel.
• Colour. An eight-band mixer for hue, saturation and brightness, colour grading for shadows, midtones and highlights, and camera calibration.
• Detail. Sharpening that leaves flat areas such as sky alone, plus noise reduction for both grain and colour speckle.
• Presence. Texture, clarity, dehaze, vibrance and saturation.
• Shape. Crop, straighten, rotate, flip and perspective correction for leaning buildings.
• Finish. Post-crop vignette and film grain.

Change one part of a picture

Darken a sky, lift a face, warm a foreground. Use a gradient, a brush, or a colour or brightness range — or let your Mac find the subject, the sky, the people or the background for you. That recognition runs on your own machine and the photograph never leaves it.

Optional help from AI, switched off until you want it

If you connect your own AI service, Lumen can suggest a starting develop for a picture, write keywords, titles, captions and alt text, offer a critique, help rate a shoot, and let you search your library by describing what you are looking for. It is off by default, asks separately before any photograph is sent, strips all metadata including location from anything it does send, and ships with no key of its own.

Export to everywhere at once

Send a finished export to a folder, a NAS, WebDAV, SFTP, Google Drive, OneDrive and S3-compatible storage in a single action. If one destination is unreachable the others still receive the file, because that is the whole point of keeping more than one copy. Location data is removed from exports unless you ask for it.

Coming from Lightroom

Lumen reads Adobe XMP sidecar files, so your ratings, colour labels, keywords and much of your existing develop work come across — and it tells you honestly which settings it could not carry over rather than approximating them silently.

Honest about what it cannot do

• No DNG or WebP export. macOS cannot write WebP, and a correct DNG needs a component Lumen does not carry. It refuses rather than hand you a file with the wrong contents inside.
• SFTP needs a key, not a password, because a password cannot be supplied safely without shipping an extra helper.
• Google Drive and OneDrive need your own client ID, because Lumen ships no credentials of its own.
• The lens profile database is a starter set. Unlisted lenses fall back to your camera maker’s correction, or to the manual sliders.

Recovery without writing to the source

• APFS write operations are not implemented. The native backend is compiled read-only and opens the source without write access.
• Existing destination files are never overwritten. Extracted data is written to a unique partial file and moved into place only after a successful read.
• Passwords stay local. A FileVault password is passed to the local backend through standard input, not exposed in command-line arguments.
• Known limits are explicit. T2 or hardware-encrypted volumes are unsupported, and the 32-bit build can access only the first APFS sub-volume.

Checking what you downloaded

The About button in the application reports the version, build date, publisher and the SHA-256 of the running executable, so it can be compared with the checksum published above without a network connection.

Three files, not a score

Point it at a repository and it writes evidence, findings and a plan. The findings name the file each conclusion came from, so you can check any of them without running it again. The plan sorts the work by severity and effort, says which items must be done before others, and gives every action a condition that proves it is finished.

It tells you what it could not check

A check with no evidence behind it reports not checked — never passed. If no vulnerability scanner is installed, the report says your exposure is unknown rather than quietly awarding a clean bill of health. Every report ends with a log of what ran and what did not, and the scorecard shows coverage next to score so a high mark on thin evidence is impossible to misread.

What it looks at

• Documentation. README substance, licence, security policy, changelog, design notes.
• Testing and CI. Whether tests exist, whether CI runs them, whether it lints and scans.
• Dependencies. Lockfiles, known advisories, automated updates, floating versions.
• Secrets and hygiene. Committed keys and .env files, large blobs in history, stale branches.
• People and change. Bus factor, commit quality, and the files that change most — where regressions actually happen.

Safe to point at anything

• It never modifies the repository it reviews. Even dependency audits run against a copy in a temporary folder, because some package managers write a lockfile into the project they are auditing.
• Nothing leaves your machine. No account, no API key, no telemetry, and no network call of its own.
• Reports are written outside the repository so they cannot be committed by accident, and secret candidates are shortened before being written to disk.

It tests, rather than asks

Most compliance tools are questionnaires with a database behind them. Vantage evaluates 49 automated tests against the actual configuration of cloud resources, identity accounts, endpoints and people — every hour, and on demand. A failing test opens a remediation task with a deadline set by its severity: three days for critical, thirty for low.

One control set, seven frameworks

159 requirements across SOC 2 Type II, ISO/IEC 27001:2022, the HIPAA Security Rule, GDPR, PCI DSS v4.0, NIST CSF 2.0 and ISO/IEC 42001 map onto a single set of 62 controls, so evidence gathered once counts everywhere it applies. Readiness is control-weighted, the way auditors actually score it.

Fixing something changes the numbers

Press Fix on a failing test and Vantage applies the compliant configuration, re-runs the test, and recomputes the control, the framework readiness and the public Trust Center in front of you. It is a working system, not a slide deck — which is why the demo is worth ten minutes.

The rest of the workflow

• Policies. 22 versioned policies with approval, annual review dates and per-person acceptance.
• People and devices. Security training, background checks, offboarding, disk encryption, screen lock, OS currency.
• Vendors and risk. Risk tiering, sub-processors, assurance reports, a residual risk heat map.
• Audit hub. Auditor details, observation windows, an evidence request list and an evidence library.
• Questionnaires. Answers drafted from your live controls and approved policies, scored for confidence, with anything below 70% flagged for a human rather than asserted.
• Trust Center. A public page generated from live monitoring, with NDA-gated documents and an access-request queue.

What the demo is, honestly

It is one shared workspace of entirely fictional data. Everyone signs in with the same published password, so anything you change is visible to everyone else until the daily reset. Do not put anything real into it — and if you want a private instance, the source is on GitHub and the server has no runtime dependencies at all.