Financial workflows
Categorization, bill entry and reconciliation across many sets of books, wired into the accounting and banking systems you already run.
Tolerance: one penny.
SysForge builds business software out of ready-made parts. You say what you need in plain words, AI puts it together from about sixty parts that already work, and a person signs off before anything runs. Every part was struck until it held its shape.
Anyone can pour molten AI into a mold and call it a product. We prefer to work the metal.
Small businesses run on duct tape: a dozen apps that do not talk to each other, and a spreadsheet everyone pretends isn’t load-bearing. Big companies hire teams to make software fit. Everyone else makes do. We start with how your work actually gets done today, build the smallest thing that helps, and harden it from there.Ready-made parts that bolt together for each job, like building blocks with the same connector on every block. Sending texts, reading documents, taking payments, keeping files, talking to AI: each is a part that already works and explains itself, so anything else on the platform can use it.
These are the ones that see the most heat. Each plate is load-rated.
Categorization, bill entry and reconciliation across many sets of books, wired into the accounting and banking systems you already run.
Tolerance: one penny.
Geographic mapping that turns addresses and boundaries into districts you can actually use.
Tolerance: precinct-level.
Radio coverage analysis built on FCC data, so the map shows where the signal really reaches.
Calibrated to FCC data, not vibes.
Open-source intelligence work, gathered and organized into something a person can read in one sitting.
Open sources only. Closed mouths.
Forms and documents filled from structured data, correctly, the first time.
Fills forms so nobody else has to.
One service in front of several AI providers, routing each job to the model that does it best.
Uses several models. Trusts none of them.
Describe the job the way you would to a new hire. SysForge drafts a plan from parts that exist and are running right now, and checks every piece fits before you see it. Nothing runs until a person says yes.
“Remind hotel guests the day before they arrive.”
An illustration of how a plan comes together, simplified for this page. The parts are real.
Five steps, always in this order. Skipping to step four is how most AI projects end up as a demo nobody uses.
Smiths read the color of steel to know how hard it has become. We read confidence the same way, and only turn the heat up when the system has earned it.
The colors and temperatures are the traditional tempering chart for carbon steel. The automation levels are ours. Most systems spend a good while in straw before anyone mentions blue.
A hallmark is the stamp that proves what the metal is. These are ours, counted from the code itself on 2 October 2026. Not a slide deck. It is running.
Four very different businesses, one set of parts. That is the point: the next industry starts with most of the work already done.
The tools a business already runs on, connected and talking both ways. More than thirty platforms so far, each one wired in for a real job, not a demo.
One connection serves many accounts: six merchants on one point-of-sale link, seven locations on one scheduling link, eight email lists on one marketing link. Build once, use everywhere.
Stripe, Square, Clover and Shopify: card payments, the till and the online store, in one place.
Sandbox first. Then the till.
QuickBooks Online and a commercial bank’s own connection, so the money and the books agree without anyone retyping.
A test login before the live one.
WebRezPro and Cloudbeds for rooms, Homebase for staff schedules, OpenTable and Cojilio for bookings.
Knows who is arriving tomorrow.
Iterable for email, BriteVerify to check an address is real first, Vista Social and Meta for social, Front and Gmail for the inbox.
Every reply comes back in, signed.
A voter data platform, an ad spending tracker and an online fundraising platform, feeding one picture of the race.
One picture, not five tabs.
Snowflake, AWS, Azure SQL, Airtable and Google Sheets, including the spreadsheet everyone pretends isn’t load-bearing.
Reads it where it sits.
BoldSign for signatures, Typeform and surveys for the questions, each answer landing where the work happens.
Signed, sealed, filed.
Anthropic’s Claude and Mistral AI, alongside models that run on our own hardware.
The right model for each job.
GitHub for every release, YouTube Live for streaming, Mapillary for street imagery, WordPress for publishing.
Ships on a signed push.
Counted from the platform’s own settings on 15 October 2026: each platform named here is set up in a service that uses it. Company names are trademarks of their owners, and naming them means we connect to them, not that they endorse us.
Everything above happens in the back. TinyKit is the part a person holds: the app on their phone, the sign-in, the screen that has to make sense at a glance. Every app we build starts from it, so each new one opens with the hard parts already done.
Eight apps run on it today, counted 15 October 2026. A fix made once reaches all of them. Same anvil, many horseshoes.
One account opens every app we make, signed in with a passkey on the phone. Nothing to remember, nothing to reset.
No passwords to leak.
What a person keeps in an app is on their phone and in their account, so every device they sign in on has it.
Changed on two phones? Both are kept.
It goes on the home screen like any app and opens without a signal, then catches up when it has one.
Works in a basement.
Large type and high contrast on purpose, every color scheme checked before anyone can use it.
Never shrinks to fit.
The phone’s own share menu, a link, or a code to scan from across the table. Files shared in land where they belong.
One tap, either direction.
Accounts and files live on machines we own and run, next to everything else, not in someone else’s cloud.
Your data stays home.
SysForge is the platform as a whole: the parts and connections in the back, TinyKit in front. The apps above are our own; customer apps on it are not shown.
Proven in the field. Now making it general, so a new industry plugs in without inheriting anyone else’s details.
Three people, one forge, a firm policy against calling anyone a ninja.
Architecture & machine learning
Swings the hammer.
Strategy & front of house
Talks to customers so the hammer doesn’t have to.
CFO, accounting & domain logic
Counts the anvils. Twice.
If any system we build ever replies with “As an AI language model,” we will take it back to the forge and re-strike it at no charge. To date, no claims. One close call.
The fine print, which is real
For early investors
We are speaking with a small first group of investors. If software built from parts makes sense to you, we would like to show you the real thing, sparks and all.