The fastest reliable way to track AI spend is one live view across every tool you pay for, plus a weekly burn-rate forecast: spend so far, divided by days elapsed, times days in the month. That matters more in 2026 because more of the bill is metered: GitHub Copilot draws down AI Credits, Cursor bills third-party models at API list price, and Claude and ChatGPT plans combine 5-hour and weekly limits with optional pay-as-you-go credits.
If you pay for a Claude plan, ChatGPT, Cursor, and some API access, each one bills separately, on its own dashboard and cycle. Most people find out what they actually spent when the invoices land, not before. Here are five ways to see it as it happens, then the forecasting method that tells you where the month will end. This post absorbed our June 2026 forecasting guide on October 23, 2026.
Option 1: The spreadsheet (free, high effort, always stale)
A manual spreadsheet is the default. It works, it is free, and it is almost always out of date. You have to log into each dashboard, copy numbers, and update it by hand. It tells you what you spent last week, not what you are spending right now, which is useless for catching a runaway session before it costs you.
Option 2: Provider dashboards (accurate, fragmented)
Every provider has a usage dashboard. They are accurate for that one provider, but you are juggling four or five browser tabs, each with a different layout and billing cycle, and none of them show your total. There is no single number for "what is my whole AI stack costing me this month."
Option 3: Cloud cost tools (powerful, overkill for individuals)
Enterprise FinOps platforms can aggregate AI spend, but they are built for organizations with cloud infrastructure and procurement teams. For a solo developer or a small team, they are expensive and far more than you need.
Option 4: Menu bar trackers (live, low effort)
A native menu bar app sits quietly in your toolbar and shows live usage across providers without browser juggling. This category exploded in 2026, CodexBar, Tally, and Tokens 4 Breakfast all live here. The differences come down to how many providers they cover, whether they track subscriptions as well as API usage, and whether they forecast future spend or just report past spend.
Option 5: Just check the bill (free, painful)
The default for most people: do nothing, and react when the invoice arrives. The problem is that by then the money is spent. Bill shock, discovering you burned $60 on Opus you did not need to, or that a forgotten subscription has been charging for months, is exactly what live tracking prevents.
How to choose
- If you only use one provider and check it occasionally, a provider dashboard is fine.
- If you use three or more tools, a menu bar tracker pays for itself the first time it catches a runaway session or a forgotten subscription.
- If you care about privacy, pick a local-first tool that does not send your usage data to a cloud server.
- If you want forecasting (not just history), make sure the tool predicts rate limits and month-end spend, most only report the past.
| Option | Cost | Effort | Live? | Shows the total across tools? |
|---|---|---|---|---|
| Spreadsheet | Free | High, manual every week | No | Yes, if you keep it updated |
| Provider dashboards | Free | Medium, one login per tool | Per tool | No |
| Cloud cost (FinOps) tools | Paid, built for organizations | High setup | Yes | Yes, for API spend |
| Menu bar trackers | Free to low one-time or subscription prices, varies by app | Low | Yes | Depends on providers covered |
| Just check the bill | Free | None | No | Only after the money is spent |
From tracking to forecasting: the burn-rate method
Tracking tells you what you have spent so far this month. Forecasting tells you where you will end up while there is still time to change course. The core method is simple: take your spend so far, divide by the days elapsed to get a daily burn rate, then multiply by the days in the month. If you have spent $30 over the first 10 days, your burn rate is $3 a day and a 30-day month ends near $90. Do it weekly, for every tool, and add the results.
- Forecast the total, not just one tool. The number that governs your budget is the sum across every provider, each on its own cycle.
- Forecast limits too. If you burn a Claude or ChatGPT weekly allowance at a steady pace, the same math tells you which day you will run out.
- Set a threshold tied to an action, for example 80% of your monthly ceiling, that triggers cheaper models, batched jobs, or pausing an agent session.
- Use per-session caps for agentic tools, so you decide the limit before a long run starts, not after the credits are gone.
FAQ
How do I estimate my monthly AI bill?
Use burn rate: spend so far divided by days elapsed gives your daily rate; multiply by the days in the month for the forecast. Do it weekly across every tool you pay for, or use a tracker that consolidates and forecasts automatically.
What is the best way to track AI API spend?
For individuals and small teams, a native menu bar tracker gives the best ratio of accuracy to effort, it shows live spend across providers without manual logging. Provider dashboards are accurate but fragmented; spreadsheets are free but always stale.
Can I track AI subscriptions and API usage in one place?
Most trackers only cover API usage. A few, like Tokens 4 Breakfast, also consolidate flat subscriptions (Claude Pro, ChatGPT Plus, Cursor) so you see your true total, not just metered API costs.
Is it safe to use a third-party AI usage tracker?
It depends on the tool. Prefer local-first apps that process data on your machine with no cloud sync and no telemetry. For Claude Code specifically, the best trackers read your local session files and do not require an API key at all.
Want more free decision tools?
The Tools hub has calculators and trackers for AI spend, plan comparison, and prompt quality.