AttoFlow AttoFlow
Trace, our first product, is in private beta

The intelligence layer for how work actually gets done.

We're building the layer that detects how work starts, moves, and finishes, end to end, inside real organisations, starting with the hours professionals bill for.

The problem

Models can't automate what they were never able to see.

Current frontier models cannot fully automate a person's workflows because they lack the granular data of how work actually gets done inside an organisation.

Today, without that data

Hours reconstructed from memory at the end of the week — half of them wrong.
Any tool granular enough to see how work happens asks you to send it to the cloud.
Automation gets designed around how work is supposed to run, not how it runs.

With AttoFlow

Every working minute captured and classified the moment it happens.
Screen-level depth that never leaves the device — by architecture, not policy.
A record of the real process — the ground truth automation has been missing.
What we're building

A record of how work really runs, built from the work itself.

We are solving that by building the intelligence layer that detects how work starts, progresses and gets completed, end to end. The first instantiation of this is Trace, which starts out as an automated timekeeping solution for professionals and organisations.

Cheapest signal first: deterministic rules resolve most activity immediately. Only what rules can't explain moves on to heuristics, then on-device AI — and every correction promotes back into a new rule.

Our first product

Trace: automated timekeeping, built on that layer.

Trace runs passively on the desktop, sees what you're working on at screen level, classifies it against your own clients and projects using a model that runs on your own machine, and turns the result into timesheets and invoices — with nothing ever leaving the device.

Passive capture. Foreground window, browser tab and input activity, sampled continuously while you work — no timers to start or stop.
One local pipeline. Capture through billing runs on your machine, and the record stays in a local, encrypted store you control.
Screen-level capture
Foreground window, browser tab, and input activity — the same depth cloud trackers collect, without the cloud.
On-device classification
A local model runs inference on your own hardware. Screen content has no network path off the machine.
Billing without a second tool
Categorised time becomes timesheets, invoices, and exports inside the same product.
Explore Trace →
12%

Anyone who bills by the hour loses about 12% of the week. Work that got done, was billable, never got invoiced.

Who it's for

Built first for people who bill by the hour.

Even in its current form, Trace is a must-have for consultants, accountants, lawyers, freelancers — anyone who deals in billable hours.

Consultants & agencies
Bill hourly across shifting client work without stopping to log every context switch.
Accounting & bookkeeping firms
Turn a team's scattered client work into accurate, exportable time by itself.
Legal teams
Screen-level capture with nothing leaving the device — built for matters that can't touch the cloud.
Freelancers & studios
Small teams that bill for their time and don't have an ops person to chase timesheets.

Trace is in private beta.

Tell us how your team bills and we'll set you up, or just ask us what we're building.

attoflow.ai@gmail.com +91 98936 60159