Know how much of your AI subscriptions you have left, before you run out.
Tally is a macOS menu bar app that tracks your Anthropic, OpenAI, OpenCode Go, and xAI subscription usage. It reads the accounts you already signed into in OpenCode, polls each provider for your current quota windows, and shows what is left and when it resets.
| Menu bar | iPhone |
|---|---|
 |
 |
You need an Apple Silicon Mac on macOS 26 or newer, and OpenCode with at least one subscription account signed in.
- Download
Tally-v0.1.0-macos-arm64.zipfrom the latest release. - Unzip it and drag
Tally.appinto~/Applicationsor/Applications. - Open it. The app is signed ad-hoc rather than notarized, so macOS may block the first launch. If it does, go to System Settings > Privacy & Security and click Open Anyway.
Tally picks up your accounts on first launch and registers itself to start at login. Nothing else to configure.
Prefer to build it yourself? See docs/INSTALLATION.md.
Click the menu bar glyph to open the popover. Tally refreshes on launch, on wake, and every two minutes. The Refresh button pulls on demand.
Manage accounts in OpenCode. Tally opens OpenCode's credential database read-only. OpenCode and its auth plugins own sign-in and token refresh. Tally uses the stored access token for usage collection, model discovery, and warm-up; it never refreshes tokens or writes to OpenCode's database. Collection works while OpenCode is stopped and stored tokens remain usable.
Switch accounts from Tally. Native and web account cards show Active in OpenCode and a Use in OpenCode button for inactive accounts. The companion supports {"action":"activate","accountId":"opaque-ID"} after resolving the account with accounts. Switching requires the local OpenCode service and uses its activation API so OpenCode reloads provider state. It changes the selected account for that provider on the Mac running Tally. Pinning and collection remain independent of selection. Selection follows Tally's inventory refresh; an unavailable inventory shows selection as unknown.
Quota bars turn red when you are burning too fast. A blue bar means your current pace fits inside the window. Red means your average usage projects that you will hit the limit before the window resets, and Tally shows how long you have. The small tick mark is the even-pace marker: where you would be if you spread the window evenly.
Pin what you care about. Pinned accounts appear at the top and their percentages show directly in the menu bar. Everything else groups by provider below. Reorder accounts in Settings, and click an account's icon to change its color.
The Activity tab counts what you actually spent. Today, yesterday, or the last 30 days of recorded OpenCode tokens and cost, broken down by provider and model, with a 30-day trend. Tally also estimates what the same activity would have cost at API rates, which is separate from what your subscription charged you.
Anthropic extra usage, OpenAI purchased credits, and banked reset credits show on the account cards. Redeeming a reset always needs an explicit confirmation from you.
Auto warm-up is off by default for every Account. Open the Warm-up tab in native Settings, turn on an eligible Account, and choose a model. Models load automatically; warming starts once a model is selected. Off accounts without an applicable five-hour window cannot be enabled and show an explanation. An existing enabled preference stays checked while waiting for quota information and can still be turned off. Menu bar pinning and ordering are in the Menu bar tab; startup and connection settings are in General. Tally schedules a short prompt after its five-hour window resets, with a random delay of up to 20 minutes and a rotating selection of 20 questions. If normal usage has already started the next window, Tally schedules after that window instead. Provider collection can add a small delay.
Web and mobile Settings have the same Warm-up controls and Menu bar section. Model lists load automatically when opening Warm-up; preferences are shared with the Mac app and follow the stored OpenCode row across token rotation. Account cards show failed warm-up attempts, and Settings has a Resume button after resolving the cause. Schedule dates use the Mac's timezone and remain visible while quota information is unavailable. Launch at login, database location, listener settings, and quitting Tally are managed on the Mac.
Tally sends HTTP requests directly to the selected provider using the selected Account's credentials. It creates no OpenCode sessions or processes and requires no executable or auth-plugin path. The model picker reads the provider's current model list.
Each warm-up reads the selected Account's current stored access token and sends one message request on the chosen model. It does not fetch the model list before sending. The selected model never falls back to another model. A failed request pauses warm-up and displays the reason in Settings. OpenCode or its auth plugin must refresh expired credentials; Tally never calls OAuth refresh endpoints or retries the message. Toggle off and on to resume after resolving an error.
Eligibility follows current quota readings, not plan names. Accounts with only weekly or monthly windows show "Not needed" and send nothing. The enabled preference remains stored so a five-hour window appearing later can resume scheduling. Missing or stale quota data shows "Waiting for quota information". Exhausted applicable allowance and pending reset operations prevent sending. Grok currently reports only a weekly window in Tally.
Warm-up runs only while Tally is running and the Mac is awake. Missed windows do not accumulate. Attempts are saved before sending; an interrupted attempt requires manual resumption. Tally sends each message once and does not replay ambiguous failures. Warm-up consumes subscription allowance and can be subject to provider restrictions; timing and prompt variation do not guarantee provider approval or prevent account suspension.
Tally serves the same interface at http://127.0.0.1:7483. To reach it from your phone, point a personal Tailscale HTTPS proxy at that port, add the exact HTTPS origin under Settings, then open it in Safari and choose Share > Add to Home Screen.
The web app follows your phone's appearance and safe areas, and shows a reconnect screen when the Mac is unreachable. Tally does not configure Tailscale for you, and no tailnet exposure is needed for local use. Never put this API on the public internet.
Tally has a REST API under /api/v1 on the same port, covering status, accounts, activity, refresh, and reset redemptions. Local clients need no token; requests are limited to loopback and the one HTTPS origin you configure.
Give your coding agent the ability to check your usage:
npx skills add MaxAnderson95/tallyThere is also an OpenCode companion plugin that registers a single tally tool for the same queries.