Measure it, cut it, and finally
understand why

Scored every 3 seconds from five behavioural signals: watch time, short-form consumption, tab switches, scroll velocity and idle time. Measured, never self-reported.

What you actually watched or read, grouped into topics — education, news, entertainment, gaming and the rest. The classification runs locally; the topic label is all that leaves your device.

Every site you visit sorted into eight fixed categories. Only the domain ever leaves your device — never the page address, never what was on it.

Focus quality split across YouTube, Facebook, Instagram, Twitter/X, Reddit, LinkedIn and TikTok, so you can see which one costs you the most and when.

Your earliest tracked sessions against your most recent, with the per-session deltas that moved: Shorts share, deep-focus minutes, tab switches.

30 days free, longer on paid plans, with CSV and JSON export of everything we hold about you.

Blocks are found automatically from sustained work — no timer to remember. 25 minutes or more counts as deep, 15–25 as focused, because recovering from one interruption takes about 23 minutes.

The site that appeared on screen when each block stopped, ranked by how often it was the one that did it.

When your deep work actually happens, across the window you pick — and the best time of day to defend, once there is enough evidence to call it.

Connect Google or Outlook Calendar read-only and see what a meeting-heavy day costs your Focus Score. Event titles and attendees are dropped on receipt — we read how many and how long, nothing else.

Thirty seconds after a session. Chips mined from what the session actually contained, so a whole entry can be one tap. Notes are encrypted — we cannot read them, which is why search runs in your browser.

Your words set against your numbers: “You wrote ‘good run’ — it was 52 minutes, no switches.” Neither a journal nor a tracker can write that line alone.

What keeps coming up across weeks of writing, pulled from your entries and nobody else's.

How the day went, in one tap and an optional line. Health data, so it sits behind its own explicit consent and its own encryption key.

Connect a wearable and see how last night's sleep tracks against today's score.

A personalised email: your trend, your worst platform, your peak hour and three things to try — grounded in your own figures rather than generic advice.

Your target comes from your own last 14 days, not a number we picked — so it is reachable on the first day and still moves. Reminders, a plan for when you will do it, and a Sunday review email that raises, lowers or keeps it in one tap.

Four GET endpoints — account, current score, today's summary, recent sessions. Authenticate with a key in an X-API-Key header, 120 requests a minute. Keys are shown once and stored only as a hash.

Point an AI assistant at your own focus data and ask it “when do I focus best?” in plain language. The Model Context Protocol is vendor-neutral, so it is not a bet on one assistant — and the four tools it exposes are read-only.

A POST when a session starts or ends, a daily score lands or a goal is met — HMAC-signed with the timestamp inside the signature, so a captured request cannot be replayed. Every attempt is logged where you would look for it.

Attach a Linear issue key to a session. The key, never the title — four hours is weaker than four hours on ENG-214, and an unbilled hour with a ticket on it is far easier to invoice.

Tag sessions to a client, set a rate, and see what your unbilled work-tool time is worth. Billing rules let you mark your own admin sites as not billable.

CSV and a full JSON export of every session, score, survey answer and account field we hold — GDPR Article 20, available from your account page without asking anyone.

SAML/OIDC single sign-on, a DPA, and data-residency options for organisations that need them.