Stats — Cost, Tokens, Sessions, and Trends
Stats is a built-in Omniscio virtual project that surfaces aggregate metrics across every Claude Code session you've run: total spend, token usage, LLM time and active time, per-project breakdowns, and trend lines over time. It lives on the Installed Plugins page (bottom of the side nav), and its In side nav switch adds it to the side nav. Before 2026-04-30 these metrics lived inside Settings → Statistics.
On this page
What it is
Stats is a built-in Omniscio virtual project that surfaces aggregate metrics across every Claude Code session you've run: total spend, token usage, LLM time and active time, per-project breakdowns, and trend lines over time. It lives on the Installed Plugins page (bottom of the side nav), and its In side nav switch adds it to the side nav.
Before 2026-04-30 these metrics lived inside Settings → Statistics. Settings is for configuration; data exploration deserves its own pane, so Stats was extracted into its own virtual project.
What it shows
These tabs, switchable from the sub-sidebar on the left:
Overview — usage value, total tokens (input/output), LLM time, active time, Your Sessions (the sessions you started or typed in, with the agent-spawned rest underneath — see "Your Sessions" below), Your Messages (just the user-typed ones), Automated Actions (combined count of in-session auto-replies + channel-driven Automations), Interruptions Avoided (a running tally of the times Omniscio handled something so you weren't interrupted — see "Interruptions Avoided" below), then a per-project table sorted by cost, then a Top Sessions list (cycle through metrics: cost, tokens, duration, etc.). The headline money tile is labeled Usage Value, not "Total Cost" — see "Usage Value vs money you actually paid" below for why. Cost values on that tile and inside the per-project table are rendered in compact
$X.XKform once they cross $1,000 (rounded to the nearest $100); below $1,000 they show as raw integer dollars (e.g.$87). Exact sub-dollar values still appear elsewhere — e.g. the "avg/session" sub-line under Usage Value.Session counts and long durations get compact formatting too:
- Session counts (the Your Sessions tile and the per-project "X sessions" badge) — below 1,000 they show as raw integers (e.g.
42); at 1,000 and above they show asX.XXXK(divided by 1,000, rounded to the nearest 0.001, trailing zeros trimmed). So2652 → 2.652K,2000 → 2K,1234 → 1.234K. - Long durations (the LLM Time / Active Time tiles and the per-project Active / LLM cells) — below 10 hours they keep the existing
Xh Ym/Xm/Xsformat (e.g.9h 59m,45s); at 10 hours and above they round to the nearest hour and add thousands separators (e.g.207h 28m → 207h,2198h 10m → 2,198h). The per-session subtitles (e.g. "avg/session") stay in the precise compact format since those values are rarely that large.
- Session counts (the Your Sessions tile and the per-project "X sessions" badge) — below 1,000 they show as raw integers (e.g.
Spend — the central "where is my AI money going?" view. One sorted list, biggest first, with a grand total at the top, combining your coding sessions (every engine — Claude, Codex, etc. — as one "Coding sessions" line) with the ~30 background AI features that log to the cost ledger (Inbox Pilot, Plain Speak, AI Suggestions, Session Titles, Context Transfer, Daily/Weekly summaries, Email Summarizer, Automations, Voice, Calendar, Knowledge Base, and more). Each row shows its dollar amount, its share of the total, and a proportional bar. Honors the Time Range selector. Engine spend is counted exactly once (see "How the Spend tab is computed" below). For most people coding sessions are the overwhelming majority — the background features are a small sliver, which the bars make obvious at a glance.
Feature Usage — counts of registered features (cron jobs created, recipes run, AI titles generated, etc.) tracked via the
feature_eventstable. A card's headline counts what a PERSON deliberately did: an action taken by the machine on its own or by an agent over the CLI keeps its own row and its own count but is left out of the total, so a background helper's total is not its polling volume. The Auto-Lander card is the clearest case — its total counts the two switch gestures, while its lands, attempts, cards, machine re-arms andcli_hitrows sit beside it. One such card is App Opens, which counts each time Omniscio is opened, split by device —mobile(a phone/tablet over web access),desktop_web(a computer browser over web access), anddesktop(the desktop app) — so you can see how often Omniscio is used on mobile vs. desktop. It counts opens (each page/app load), not time-in-app, and records only the device bucket, never the browser identity.Inbox — Inbox Attention Analytics: a 100% local report on how you triage the inbox — average time-to-clear, average items waiting (event-sourced depth), looked-but-didn't-act, and per-item view time, broken down by type / project / session, plus a per-day depth strip and an actions-taken breakdown. Honors the Time Range selector; on by default (Settings → Diagnostics toggle). See inbox-analytics.md.
Shortcut Efficiency — a passive stats tab surfacing keyboard-shortcut adoption. See shortcut-efficiency.md.
Trends — three line charts over time: Sessions per Day, Cost per Day, and Tokens per Day. The tokens chart plots each local day's real usage — the session tokens used that day plus background-feature tokens — and carries a one-sentence note that the per-day ledger began on 4 October 2026 with no backfill, so a session that ran before then has all of its tokens counted on its start day.
Usage — the four-chart rate-limit forecast (per-account usage rates, projected exhaustion, hour-of-day model). See usage-forecast.md.
Records — a gamified "trophy case" of ~18 all-time personal bests (longest session, priciest day, longest streak, lifetime token/session tiers, automation leverage, and a Peak Parallelism family) that celebrate with tier badges + confetti when broken. See records.md.
Mission Control — listed right after Records, and only while Mission Control is switched on. It opens a page with sub-tabs, and remembers the one you last used:
- Overview — across all your projects: On track and At risk counts taken straight from the risk heatmap's own ratings (no separate "on track" rule), short notes for projects that don't have enough history yet or have nothing due, and Team pace — items finished per day across every workspace of your account over the last 28 days. Two buttons jump to Projects and Omni (See Omni charts).
- Projects — Project Analytics for one board at a time (pick it at the top): intelligence snapshots, Velocity History, the deadline forecast (the simulation-based one, p50/p85/p95), sprint history, lead time and the seasonal calendar.
- Omni — the same risk heatmap Omni's Operations view shows. With Omni off, it says "Turn on Omni to see its charts here." and offers an Open Omni button. The tab shows the custom name if you gave Omni one.
- Developer Analytics — developer builds only: counts of which Mission Control features get clicked. Its numbers follow the Time Range, so the picker shows only while this sub-tab is open; its "Tracking since" date carries the year and follows your UI language. An installed copy never shows it.
- With no boards, every sub-tab says Create a board to see analytics. A saved Project Analytics view from before reopens on Projects, and a developer's saved Mission Control usage view on Developer Analytics.
- On a phone or the web, Projects shows the velocity trend and deadline forecast for one picked project, with no Developer Analytics. Only your member projects in owned workspaces appear.
Developer-build-only tabs — Settings Search and Reply Method are product-team telemetry readouts, so they appear only on a developer build (running from source). An installed copy of Omniscio never lists them, never reopens one that was saved as the last tab (Stats opens on Overview instead), and never warms their data in the background. See settings-search-telemetry.md and reply-method.md.
Nudges — retired, so it never appears: Feature Discovery Nudges was switched off for every user on 2026-08-31 and its Settings → Lab toggle was dropped, so the tab has no install that can show it. While the feature was live, the tab put the pitched feature's usage before the nudge next to after it, plus the shown → accepted/dismissed funnel. See feature-discovery-nudges.md.
The sub-sidebar also has a Time Range selector that filters Overview, Spend, and Feature Usage:
- Today — sessions/events from local midnight to now.
- Yesterday — sessions/events bounded between midnight yesterday and midnight today (the only range with both a lower and an upper bound).
Today and Yesterday count the work done that day. Their money (Usage Value, the Spend tab, per-engine and per-project costs, top sessions) and their token totals add up what each session actually used on that day, read from the same per-day record the Daily Digest uses — so Statistics' Yesterday and that day's Digest show the same figures. A long-running session that started days ago still counts for the day it worked, and a session that keeps working after midnight does not keep growing yesterday's numbers. The Sessions count and the message counts still mean "sessions that started that day". The rolling ranges (7, 30, 90 Days) still add up each session's whole lifetime by its start time. The one place Yesterday can differ from the Digest is money from a session you deleted: Statistics keeps counting money that was spent, the Digest leaves it out.
- 7 Days / 30 Days / 90 Days — rolling windows ending now.
- All Time — no filter.
Picking a range shows that it is loading. The numbers already on screen stay put until the new range's numbers arrive, so a slow load never flashes an empty page — but they dim and a spinner appears right away, on every tab, and both clear the moment the new numbers land. Before 2026-11-11 nothing on screen changed while the new range loaded, so on a busy computer a click looked ignored. Each tab's own labels keep naming the range of the numbers actually shown (for example "Usage Value (7 Days)" while 30 Days is still loading).
"Today" and "Yesterday" use your local-day boundaries (the JS setHours(0,0,0,0) of the current Date, then converted to UTC for the SQL filter), so the labels match the day shown on your wall clock.
Which day money lands on. Spend is counted on the day it was actually spent, not the day its session started. Every money figure for Yesterday, 7 Days, 30 Days and 90 Days (Spend, Overview's cost, the per-project cost column, top sessions by cost) and every bar of the Trends Cost per Day chart reads the per-day cost ledger (session_daily_usage, written on every charge since 2026-10-04): a session that ran across five days puts each day's share on that day, and a long-runner started before the window still counts for what it spent inside it. For 7, 30 and 90 Days, a session that started inside the range also keeps any spend the ledger never saw (in practice, money spent before the ledger existed) on its start day; Yesterday is the ledger's own rows for that day. 7 Days / 30 Days / 90 Days count their oldest calendar day whole, so just after midnight "7 Days" covers the full day a week back plus today so far. The per-engine drill-down's token counts for these ranges come from the same ledger, so a long-running session shows the tokens it used inside the range, not its whole life. Today reads the same day-delta it always has. Before 2026-10-14 these ranges put each session's whole lifetime cost on the day it started and dropped sessions that started earlier: on the owner's install "Yesterday" (2026-10-13) read $9,436.73 against $13,647.96 actually spent, and the Oct 6 Trends bar read $23,670 against $11,292.
The Trends tab uses the same Time Range selector but treats it as a window-length anchored on today: Today = 1 day, Yesterday = 2 days (yesterday + today, since the trends endpoint always anchors to today), 7d/30d/90d as expected, and All Time maps to 365 days (the existing one-year cap on the trends endpoint).
Which day tokens land on. The Tokens per Day chart reads the same per-day ledger the cost chart does: each bar is the input plus output tokens used on that local day, so a session that ran across five days puts each day's share on its own bar, and a session that started before the window still counts for what it used inside it. A session that started inside the window but has usage the ledger never saw (in practice, everything from before the ledger shipped on 2026-10-04) keeps that remainder on its start day. Background AI features are added from the charge log, with the session engines excluded so a token is never counted twice, and cache reads/writes never inflate a bar — they are already inside the input total. The series is zero-filled across the whole window, so a quiet day reads as a real zero rather than a gap.
The Records tab is all-time and ignores the Time Range selector entirely — the control is hidden when Records is active (a "most sessions ever in flight at once" trophy doesn't make sense scoped to "Today"). The Usage forecast tab likewise doesn't use the Time Range selector.
Note: For the Feature Usage tab, per-item features (such as snippet usage counts and automation run counts) only have an all-time total — there is no day-by-day history for them. When you pick a non-"All Time" range, those features show 0; events-sourced features (gmail, snooze, etc.) honor the filter normally.
A refresh button (top-right of the sub-sidebar) re-fetches the data. It spins only for a refresh you asked for — the background warm below never touches it.
Tabs load ahead of you. Opening Stats starts quietly loading every other tab's data in the background, so switching tabs is instant instead of showing skeletons while a fresh query runs. The warm is deliberately unhurried — it waits for an idle moment and loads one tab at a time — so it can never slow down the tab you're actually reading. Changing the Time Range starts a fresh warm for the new range, and leaving Stats stops it. Whatever hasn't warmed yet simply loads normally when you click it, exactly as before.
Where to find it
- Open Omniscio.
- Open Installed Plugins at the bottom of the side nav.
- Click Stats (chart icon).
The Stats panel opens. There is no separate modal anymore — the Settings → Statistics tab is gone.
How it behaves
Usage Value vs money you actually paid
The big dollar figure at the top of Overview (and the "All AI usage" card on the Spend tab) is API-equivalent usage value, not a bill. It answers "what would all this AI work have cost at metered API rates?" If you run your coding sessions on a subscription account (a Claude plan you signed into, a ChatGPT plan behind Codex, etc.), that work is already covered by the flat monthly fee — the dollar figure is what it would have cost, not what you paid. That's why a heavy user can see a five-figure Usage Value while having paid only their subscription. The number is genuinely useful (it's how you know your plan is paying for itself), but it is not money leaving your pocket.
The out-of-pocket number is the honest "money you were actually billed" subset: charges made against a real API key (your own Anthropic/OpenAI/etc. key, billed per token) rather than a signed-in subscription plan. You see it in two places:
- Overview → Usage Value tile — when part of the total is real out-of-pocket money, the sub-line reads "$X paid out of pocket, the rest covered by your plans." When there's no out-of-pocket spend to distinguish, it shows the usual avg/session sub-line.
- Spend tab — a dedicated Out of pocket card next to the "All AI usage" grand total, with the sub-line "Billed to your own API keys, not covered by a plan."
How the split is classified: session spend counts as out of pocket when the session ran on a real API-key account, OR on an own-key metered vendor engine (DeepSeek, Kimi, GLM, MiniMax, Meta, CrofAI) — those bill your real per-token vendor key even though they carry no signed-in account, so they're real money. Sessions on signed-in subscription accounts are plan-covered, as are non-metered sessions with no recorded account (a Codex run on a ChatGPT plan, Gemini, and the like). Background feature charges in the cost ledger count as out of pocket unless they are Anthropic charges made by a signed-in Claude subscription account — a DeepSeek, OpenRouter or DeepInfra charge is a real vendor bill even when a Claude login account triggered it (before 2026-10-14 those were wrongly counted as plan-covered: $16.30 over 30 days on the owner's install). Spend preserved from purged sessions keeps its original classification; purged background-feature spend can no longer be classified and is counted plan-covered. The out-of-pocket figure is always between $0 and the total.
Note on freshness: the message counts (Your Messages / Automated Actions, and the per-project counts) are kept as a per-session running tally that a background task refreshes shortly after activity, so the Overview loads instantly even across a very large history instead of re-scanning every message. When the definition of a count changes, the app clears every session's tally, so the Overview falls back to reading the messages directly — correct straight away, just slower — while the background task refills it. That is how the 2026-09-09 "Your Messages" correction reaches an existing install, with no separate backfill step to run. In practice the numbers are current; during heavy, continuous chatting a count can lag by up to about a minute before it catches up. The Your Sessions count reads the same tally to tell whether you typed in a session, so it can lag the same way. Every other metric (cost, tokens, time, the total session count) is exact and immediate.
Your Sessions — yours vs agent-spawned
The Sessions tile on Overview headlines Your Sessions, not every session row. Most sessions on a busy install are started by other agents — code and plan reviewers, crew memory scribes, cloud runs, fixers — so a raw total badly overstates your own work (on the owner's install, 2,864 sessions started on 2026-10-13, and 111 of them were his).
- Counted as yours — a session you started from a place where you type the opening message (the main chat, Quick Launch, the mobile new-session screen, Team Chat, Supermail, Clean Room, the PDF viewer, Ask about this page, search's Ask Claude, a re-run of your own message, any session host's + New, or a session with no recorded launch source — the app's default in-app launch, the same "Me" rule as the Ctrl+K Started by filter), OR any session you typed at least one message into, even if an agent started it.
- Not counted as yours — sessions an agent, automation, cron job or the cloud started and you never typed in. Tapping an inbox card's Start session counts only once you type in that session, because the app wrote its opening prompt.
- The line underneath reads
+N agent-spawned · M turns: how many other sessions started in the range, and total turns across all of them. Both numbers come from one backend read, so yours plus agent-spawned always equals the total, on desktop and on a phone alike. - The empty-range and no-data screens still key on the total, so a range with only agent sessions shows the tile at 0 rather than an empty state.
How the message counts split
The Overview tab shows two separate cards that summarize what got sent in your sessions:
- Your Messages — messages you typed yourself in the composer. Aside replies (the small follow-up questions you can fire alongside a session) count here too because you typed them. The sub-line shows the average per session you messaged. It deliberately does NOT count five things that land in the same place your messages do: a message another agent sent this one, a message another session or a scheduled job pushed in over the CLI, a board notice (another agent telling this session that something was posted — T-176), a row the CLI queued without naming a sender, one of the app's own automatic nudges to an agent (a friction-fix reminder, or a to-do reminder at a pipeline phase change — added 2026-10-09), or a session's opening prompt (the Your Sessions card already counts a session you opened from a place where you type, and when a machine started the session that prompt is a machine's words). On a real install, board notices and those nudges together were 4,886 of the 5,360 messages the briefing credited for 2026-10-08. Before 2026-09-09 all three counted then, and the board notice and the sender-less CLI row went on counting until 2026-10-08, so the card badly overstated you — on a real install it read 10,700 against a real 4,645, and the daily briefing's "answered" chip showed 2,426 for a single day of which 73% was machinery. One shared definition now backs both surfaces (human-operator-turn-sql.ts), so the two screens can never disagree about who "you" is. Total Tokens — the "in" figure includes cache re-reads (context the model read again), which is most of it on a Claude Code day, and the sub-line says so. Avg Response Time uses the same definition (since 2026-10-15): it times only replies you wrote, so agent-to-agent turns no longer pull it toward zero, and it reads every such reply inside the selected range (for All Time, the 500 most recently started sessions you messaged). Before, it sampled the 500 most recently started sessions of any kind — about four hours on a busy install — and 70 of its 106 samples were machine-written. The sub-line averages divide by the sessions you messaged in the range, not by every session the agents ran.
- Automated Actions — combined count of two things Omniscio did on your behalf without you typing:
- In-session auto-replies: messages sent into a Claude session by an Away Mode auto-reply rule (the "When the agent says X, reply with Y" rules under the Automations virtual project's AUTO-REPLIES category, sometimes called Quick-Reply rules). Counted from
conversation_messagesrows whosemetadata.autoRulesnapshot the engine attaches at write time; rows without that snapshot are excluded as ambiguous. - Channel-driven Automations: actions taken by your Automations rules outside the session conversation — SMS auto-replies, Slack/Telegram forwards, Gmail summaries, archive actions, label moves, notify pings, etc. Counted from
automation_runsrows where the rule actually matched the message (condition_matched = 1), the action succeeded (action_result = 'success'), and the action was something other than the explicit no-oppassaction (which exists so a chain can match without any side effect). - When the combined count is 0, the sub-line reads "No automated actions taken." Otherwise it shows the breakdown ("X in-session, Y channel-driven") so you can tell which side dominated.
- In-session auto-replies: messages sent into a Claude session by an Away Mode auto-reply rule (the "When the agent says X, reply with Y" rules under the Automations virtual project's AUTO-REPLIES category, sometimes called Quick-Reply rules). Counted from
- System plumbing — counted in neither of these two cards. When Omniscio nudges a stuck session (rate-limit recovery, "Please continue" auto-continue, auth retry, stall watchdog, silent-turn auto-retry, crash-recovery nudge, session-restart recovery), it sends a message under your identity but the message is plumbing, not a real reply. It's excluded from the two message-count cards above so they stay an honest answer to "how much did I write?" and "how much automation ran?". (A few of these — the "Please continue" auto-continue and the background-wait hold — DO count in the separate Interruptions Avoided card below, which measures a different thing: moments you weren't interrupted, not messages.)
Channel-driven actions are not broken down per-project on the By Project table because they live outside the session/project graph.
Interruptions Avoided
The Interruptions Avoided card (Overview tab) is a running tally of the moments Omniscio handled something autonomously so you did not have to look, click, read, decide, or reply. It answers "how much is Omniscio saving me from being interrupted?" The headline number is the sum of three buckets, shown in the sub-line ("N self-archived, M auto-replied, K auto-approved"); when it's 0 the sub-line reads "Nothing handled automatically yet." Hovering the info (ⓘ) icon next to the card title shows a one-line plain-English explanation of each bucket.
- Self-archived — a session finished its work and archived itself off your board (via the
[[OMNISCIO_SELF_ARCHIVE]]marker or the self-archive route). Counted only when the archive actually applied immediately — a self-archive that's queued for your approval (because you require approval for agent lifecycle actions) does not count, since you still had to act. - Auto-replied — the app replied or held on your behalf instead of flipping the session to "Needs You" and pinging you: an Away-Mode auto-reply rule fired, an auto-continue nudged a mid-task agent onward, a re-woken session found nothing to do and quietly went back to sleep, or a session declared it was waiting on a background process and Omniscio kept it running.
- Auto-approved — a dev-pipeline gate advanced without you having to approve it.
It honors the Time Range selector, so "All Time" is your lifetime running tally and shorter ranges show a recent slice (attributed by each session's start time, like the other session metrics). Unlike the message counts above, this number is exact and immediate — it's a direct tally kept on each session, not the background-refreshed message count.
Overlap with Automated Actions is intentional. An Away-Mode auto-reply counts here and in the Automated Actions card — they answer different questions ("how many times was I not interrupted?" vs. "how much automation ran?"), so the same underlying event legitimately appears in both.
Turning it off. Settings → System → Show interruptions avoided hides the card (default on). The counting itself always runs in the background (it's near-free), so turning the card back on shows the full tally, including any time it was hidden. Like the rest of Stats, it has no CLI route — it's a read-only renderer surface backed by the local database.
What the By Project table shows
Each row is one project; rows are sorted by cost descending and capped by the same time-range filter as the cards above. A thin accent bar under the project name shows that project's cost relative to the top project. Four stats per row:
- Cost — total spend for sessions in this project, rendered in compact
$X.XKform once it crosses $1,000 (e.g.$1.2K); below $1,000 it shows as raw integer dollars (e.g.$87). - Tokens — combined input + output tokens (e.g.
1.5M). - Active — accumulated engaged seconds across this project's sessions (see "How Active Time is calculated"). Below 10 hours it shows as
Xh Ym; at 10 hours and above it rounds to the nearest hour (e.g.207h). - LLM — time the LLM was actually thinking, summed from
sessions.duration_api_msacross this project's sessions. Hovering reveals a tooltip: "Time Claude spent thinking (sum of LLM API call durations)." This is the per-project counterpart to the top LLM Time tile. Same long-duration formatting as Active —2,198hat 1,000+ hours.
There is no longer a per-project "Your msgs/sess" cell — the cards above already summarize human vs. automated message counts globally, and the per-row column proved confusing without adding clear signal.
How Active Time is calculated
"Active Time" is the only metric on the Overview tab that Omniscio infers rather than reads from the model — it's an estimate of how long you were actually focused on a session, not how long the session sat open.
The renderer samples engagement on a 5-second tick. A tick counts as engaged when both are true:
- The Omniscio window is focused (Electron
document.hasFocus(), or for the mobile/web PWA the browser tab is visible). - You produced an input event in the last 2 minutes —
mousemove,keydown,click,touchstart, orscroll.touchstartis what makes mobile tap-and-long-press count even when no mouse fires; tap also generates a syntheticclickand touch-scrolling firesscroll, so both desktop and mobile are covered without a separate code path.
Every 30 seconds the accumulated engaged seconds are flushed to the main process and added to the currently-viewed session's engaged_seconds column. Sessions are bucketed into the time range by their started_at timestamp — so engagement you spend today on a session that started yesterday lands in Yesterday's bucket on the Stats page, not Today's.
Failure recovery: if the heartbeat IPC drops (transient WebSocket loss for the mobile PWA, or the server rejects the envelope), the unflushed seconds are restored to the next heartbeat instead of being silently lost. The buffer is capped at 120 seconds so a long offline stretch can't grow it without bound; engagement beyond ~4 minutes of continuous failed flushes is dropped (preferable to retrying a payload the server will always reject).
What this means for the number you see:
- Time you spent reading without scrolling, mousing, or typing for >2 minutes drops out (idle threshold) — even though you may have been actively reading.
- Time on the inbox or other virtual projects (no attributable session) is not added to any session's
engaged_secondsand so doesn't show up in Active Time at all. - Time on a session in another window/app does not count — even if Omniscio is open, focus has to be on Omniscio.
If the number feels low, the most common causes are reading-without-input (idle dropout) and time spent on virtual-project views.
The same 30-second heartbeat also records app-wide focused time — every engaged second on any page, inbox and other virtual projects included — into a small per-minute local table (app_focus_minutes, capped at 60 seconds per minute and kept 14 days). It is not shown on the Stats page; it rides the anonymous daily fleet digest and, for a signed-in user, the daily activity record as focusedHours, so the admin console can report how long people really spend in the app. The per-session Active Time above is unchanged.
How the Spend tab is computed
The Spend tab answers "where is my AI money going?" by merging two sources of cost that Omniscio already records. Two headline cards sit at the top: All AI usage (the grand API-equivalent total for the selected range) and Out of pocket (the subset actually billed to your own API keys — see "Usage Value vs money you actually paid" above). The sources:
- Coding sessions — the sum of
sessions.cost_usdacross every session (all engines combined into one line). This spend is immortal: every dollar you've ever spent counts, even on sessions you later deleted, sessions whose whole project you deleted, and sessions old enough that retention has permanently purged them (their cost is snapshotted into a tiny durable ledger just before deletion, so it's never lost). Deleting things never retroactively shrinks your spend history — it works like the background-feature ledger, which no delete can touch. The per-account cost views in Settings → Accounts count spend the same way, so the numbers stay reconcilable. - Background features — per-feature spend from the
api_cost_logcost ledger, where each row'ssourcenames the feature. Many internal source codes collapse to one feature name (every Plain Speak pipeline stage → "Plain Speak"; the routing + overlay decisions → "Inbox Pilot"; both title generators → "Session Titles"), so you see one row per feature, not per internal stage. A feature without a friendly name yet shows a tidied-up version of its raw code rather than disappearing. - Tasks sessions count under Tasks — a session Tasks launched (a row or batch agent, a Break it down, a close-out action, a plate plan, a caught-item triage, or the Tasks session host) is moved out of "Coding sessions" and into the "Tasks" row, so that row answers what the whole Tasks feature costs rather than only its small background AI calls. The total is unchanged and still equals the sum of the rows: the money moved between rows, it did not go down. "Spend by engine" under "Coding sessions" leaves those Tasks sessions out, so it still adds up to the (now smaller) Coding-sessions row. The row's expanded detail still lists only the itemized background calls, because a session's cost is one per-session figure rather than itemized charges; and a Tasks session that retention has already purged falls back into Coding sessions, because the purge ledger keeps the dollars but not which feature started them.
Counted once — no double-counting. External engines (Codex, Pi, and the one-shot runner) record their cost in two places: on the session and in the cost ledger under the engine's name. The Spend tab drops those engine rows from the feature breakdown — their dollars are already in the "Coding sessions" line — so the grand total is honest. The total shown always equals the sum of the rows below it.
Spend by engine. Beneath the "Coding sessions" line, a per-engine drill-down splits that same session spend across the models you actually ran — Claude, Codex, Pi, Gemini, Cursor, and so on — biggest first. It always adds up to the "Coding sessions" total (it's that one number, partitioned by engine). Two honesty markers:
- "≈ estimated" — Codex reports how many tokens it used but not a dollar figure (it bills against a ChatGPT plan, not a metered API), so Omniscio prices those tokens at OpenAI's published rates. Gemini and Cursor are estimated the same way, priced at the chosen model's published rates. A close estimate, marked with a "≈".
- "cost not reported" — some engines (Antigravity, Hermes, OpenClaw) don't report cost or token usage today, so Omniscio can't compute what they cost. Rather than a misleading $0 (which reads as "free"), they're listed as "cost not reported by the tool" — they ran, the cost just isn't measurable yet. Engines that do report real cost (Claude, Pi, OpenCode) show their exact dollar amount.
The same "≈" marker appears next to a single non-Claude session's own cost, so an estimated figure always reads as an estimate.
Spend by model, and your own prices. Below the engine split, Spend by model groups your coding sessions by the model they ran on (sessions that never recorded a model show as "Default model"), with each model's tokens — fresh input, cached, output — and recorded cost. Press Your prices to set, per model, your own price in US dollars per million tokens (input, output, and optionally cache read and cache write — a blank cache price uses the input price, 0 means free), or to price a model as another model (handy for a custom or locally named model id, or a gateway id the price list does not know). Each priced row then shows At your prices $X, and a line above the list compares the recorded total with the total at your prices. Your prices only change these "at your prices" figures: the recorded spend, the totals above, and every spending limit keep using the recorded figures. Sessions from before the app started recording the cache split (September 2026) stay at their recorded cost inside the "at your prices" figure, and the row says how many there are. The list shows the top 50 models; any beyond that are named in a note and still counted in the totals. Adapted from T3 Code's usage page (MIT).
How much was re-read from cache. Each engine row also shows the share of its input tokens that the model re-read from its own prompt cache ("94% cached"), and, when that engine ran more than one model, a line per model with its own share underneath. This is the answer to "am I paying to re-send the same context over and over": a coding session's input is dominated by context the provider already has, and a cache hit is billed at a fraction of the full input rate, so a low share is the one number worth looking at. Hovering any figure names the counts behind it ("1,250,000 of 1,330,000 input tokens were cache reads"), because the percentage is a share of tokens, not a saving.
Two rules keep the figure honest. A range that predates the cache accounting (Claude sessions before 11 September 2026, or an engine that has never reported a cache split) shows an em dash rather than a fabricated 0% — and those old tokens are left out of the percentage entirely, so a long history never drags today's share down. The share counts cache reads only; tokens written into the cache on the first pass are new material and sit in the denominator with the rest of the input.
Claude costs from before 7 September 2026 are a corrected estimate. Until that date Omniscio recorded a Claude session's cost by adding up the running total Claude Code reports after every turn, so a long session's earlier turns were counted again and again and old figures read roughly five times too high. A one-time correction (2026-10-16) replaced each of those inflated figures with an estimate: the session's input tokens times the cost per million input tokens that real, correctly recorded Claude sessions showed in the two weeks after the fix (a separate rate per session length, or per model where the session recorded one). It only ever lowers a figure, never raises one; it leaves Codex and other engines alone, as well as sessions that were already rebuilt exactly from their transcripts on 2026-09-09, sessions with no recorded tokens, and sessions started after the fix. The original figures are kept in the database table claude_cost_history_estimate_backup, and the Records tab's Big Spender and Biggest Spend Day were lowered to match. The transcripts behind these sessions no longer exist, so the corrected figures are a best estimate, not an exact bill.
Corrected Codex history (2026-10-16). Before the fix that tracks each Codex thread separately, a Codex session whose sub-agents ran alongside it could be charged the gap between two threads' running totals over and over. A one-time update lowered those charges to what Codex's own per-thread records show was really used. It only applies where every charge was matched to those records, only lowers and never raises, and keeps the original numbers in a backup table. On the owner's computer that took one session from $2,930.04 to $474.20 and the Codex total from $6,001.95 to $3,546.11; on any other computer it changes nothing, and sessions the same bug under-counted were left as they were.
Expand a feature row (drill-down). Beside the per-engine split above, every feature row (Email Summarizer, Session Titles, Plain Speak, Aside, …) expands to show the individual charges behind it: a per-model split, the charge count, and the biggest individual charges, biggest first. Each charge is resolved to a friendly label where Omniscio knows one — the session it belongs to (by name), a short context label (an email subject, a document name…), or, for Asides, the question you asked; charges Omniscio can't attribute show just the model and time. The list is capped at the biggest few (with an honest "Showing the N biggest of M charges." line when there are more) and it honors the Time Range selector like the rest of the tab. Attribution is forward-only: a charge carries its session/label only from the point that feature started recording it, so older charges drill down to model + time only. It all stays on your machine — the session id and label are local metadata, never sent anywhere.
Keyboard shortcuts
| Key | What it does |
|---|---|
| Alt+Shift+U | Open Stats, on the tab and time range you last used (works from anywhere, not while typing in a text field) |
| Alt+Shift+C | Show the Spend tab |
| Alt+Shift+X | Show tokens — the Overview tab with the token totals |
| Alt+Shift+Z | Show limits — the Usage tab with the rate-limit forecast |
| Alt+Shift+1 / 2 / 3 / 4 | Set the Time Range to Today / 7 Days / 30 Days / 90 Days |
The tab and range keys work only while Stats is the screen you are looking at (on a phone with a keyboard, the Stats list or a Stats tab), and each does exactly what clicking that tab or range does. Today is today's calendar day, not the last 24 hours. On a tab without a time range (Usage, Trends, Records …) a range key changes nothing and says so. Change any of them in Settings → Keyboard Shortcuts. The key set follows T3 Code's usage-page shortcuts (MIT).
Your other computers
When you link your own computers (Other computers), Stats adds up the sessions every one of them ran: the totals, the yours-vs-agent split, the averages, the daily charts, the By Project table and the Spend tab, where they appear as one Other computers row. Each session counts once, on the computer that ran it. Settings → Remote Access → Link my computers → Stats from my other computers switches between Combined (the default), Combined, with a per-computer breakdown (a By computer section on the Overview, each with the time its numbers were collected), and This computer only. The Inbox tab and the per-engine drill-down stay this computer's own. If the other computers' numbers cannot be read, Stats shows this computer's own numbers instead of an error.
Privacy
Per-session metrics are stored locally. The "By Project" breakdown shows folder paths you've already exposed to Omniscio by registering the project. Nothing here is shared externally.
Related
- Settings — for configuration (auto-replies, accounts, themes, etc.)
Feature events architecture(repo-only) — how thefeature_eventstable is populated
Last verified 2026-11-16