Marling

How Marling works

A private AI utility for your office, governed on-site, tuned for your profession.

Marling is a quiet mini-PC that sits on your network. It puts one friendly chat window in front of the best AI models, runs privacy and spending controls on the box itself, and arrives already configured for the way your practice works. Here's what's inside.

Privacy & data handling

Your data stays on your terms.

Rolling out for regulated practices

On-box de-identification

Before any message reaches a cloud model, Marling detects identifiers (names, Social Security numbers, medical-record numbers, phone numbers, email addresses) and replaces them with entity-type placeholders, right on the appliance. The cloud only ever sees the masked version. Your finished draft comes back with the real details restored, on the box. It's reversible, so your draft stays personalized, and fail-closed, so if the masking can't run, the message doesn't send.

You type: stays on the box

Patient John Smith, SSN 123-45-6789, MRN 88421, phone 555-123-4567

What the cloud model receives

Patient <PERSON>, SSN <US_SSN>, MRN <MEDICAL_RECORD_NUMBER>, phone <PHONE_NUMBER>

on-box de-identification + minimum necessary + your BAAs

This is defense-in-depth and minimum-necessary handling, not HHS Safe-Harbor de-identification, and not a substitute for your own Business Associate Agreements. Automated detection can miss identifiers buried in free text. We'll walk you through exactly what it does, and doesn't do, before you go live with real records.

  • On-box routing

    Marling decides which model best fits each message using a local model on the box, so even the routing decision never leaves the building.

  • Fully-local mode

    Run an on-box model for air-gapped conversations with no cloud involved at all, when a task calls for zero data leaving the office.

  • Private by default

    No public ports. The box is reachable only over a private encrypted network, never the open internet.

Built for non-technical staff

It just works. No prompt expertise required.

Profession-tuned templates

A template is a one-click shortcut for a document the office writes constantly. Instead of explaining to a chatbot what the document is every time, your staff pick the template, fill in a few blanks, and get a properly structured draft in your profession's format. The know-how for how the document should be built is already baked in, so the newest hire produces the same quality as the most senior.

Each pack ships dozens: SOAP and progress notes, intake summaries, referral and demand letters, parenting plans, financial-affidavit covers, payment reminders, review requests, work-order summaries, and more.

/demand-letter

  • On behalf of[client]
  • Recipient[name / entity]
  • Facts[what happened]
  • Relief demanded[amount / action]
  • Deadline[date]
  • Jurisdiction[governing law]

→ a structured demand letter in the firm's format. Brackets anything missing. Invents no facts or citations.

  • Lands on the right assistant

    Staff open to a single curated assistant with prompt suggestions for their work, not a confusing list of models to choose from.

  • Searches the web on its own

    When the assistant needs current information it searches automatically (nothing to toggle) and trims results to stay fast and within budget. (Rolling out.)

  • Plug-and-go setup

    Connect the box, open a browser, “Set up your assistant,” and you're working. No installation, no IT visit.

One office, one box

A shared appliance, not a pile of subscriptions.

Marling is one box the whole practice shares, like the file server or the office printer, except it makes everyone's writing faster. Every staff member gets their own login to the same configured assistant and template library.

  • The same tuned tool for every hire

    New paralegal, new front-desk, new associate: all sit down to the identical assistant and templates. No tribal knowledge about which prompts to use.

  • Roles built in

    Clinical staff get the full clinical toolset; front-desk staff get only the non-clinical assistant. Enforced on the box, not just suggested.

  • Owner-managed, no IT

    The practice owner adds staff and resets passwords from one simple page. No procurement, no per-seat licenses, no admin internals to touch.

  • Improve once, everyone benefits

    Templates live on the box. Tune one and the whole office's output improves at the same time.

Cost control & peace of mind

A cost that can't run away.

Per-seat AI scales with both headcount and how heavily people use it. Marling is a flat box with a hard daily ceiling you set, so your total AI cost stays bounded and predictable no matter how busy the office gets.

  • Hard daily spend cap

    Set a daily budget ceiling on the box's AI spend. When it's reached, that's the limit. A runaway bill is impossible by design.

  • Spike & budget alerts

    An instant notification if spending spikes or a cap is approached, so there's never a surprise at the end of the month.

  • Every request logged

    Usage is recorded to the cent, so you always know exactly where you stand.

  • Fleet health at a glance

    For multi-box practices, see the health of every appliance and get an immediate alert if one goes offline. Metadata only, no client data in the telemetry.

Reliability

Built to run unattended.

  • Self-healing on reboot

    After a power cycle, the box restores its assistant automatically. It's ready when staff arrive, with nothing to restart.

  • Reproducible provisioning

    Every box is built from the same tooling, so they're identical and can be updated in the field without a site visit.

See it in your practice.

We're onboarding our first ten founding practices now. Tell us about yours and we'll send the details.

Get early access →

Founding pricing for the first ten practices.