// SHIPS MAY 12 · BEFORE GITHUB'S OWN PREVIEW TOOL

Know your Copilot bill before GitHub does.

GitHub Copilot switches to token-based billing on June 1, 2026. Septim Meter is a one-time $29 CLI + HTML report that projects your monthly bill, breaks down the math by workflow, and shows you when calling Anthropic / OpenAI directly is cheaper than staying on Copilot. Self-hosted, runs offline on your laptop, no API key required.

Countdown: 34 days until GitHub's June 1 billing change. Septim Meter ships May 12 — 20 days inside the window.

The change in plain English.

On April 27, 2026 GitHub announced that Premium Request Units are being replaced by AI Credits. 1 AI credit = $0.01 USD. Plan prices stay the same: Pro $10, Pro+ $39, Business $19/user, Enterprise $39/user. Every plan includes its dollar value worth of credits ($10 = 1,000 credits on Pro).

Code completions and Next-Edit suggestions stay free. Everything else — chat, agentic sessions, code review — is metered against your credit pool at the published per-million-token rate of whichever model you used. Heavy agentic users on Sonnet or Opus will spend their entire monthly credit pool in days. GitHub said a preview cost tool is "coming in the coming weeks." It does not exist as of April 28.

What your June bill probably looks like.

Septim Meter ships with three calibrated profiles you can run in 60 seconds. Numbers below come from the validated formula module — the same math GitHub uses, sourced from docs.github.com/copilot-billing.

// LIGHT USER
~150 chats/mo on GPT-5 mini, no agentic
$10.05/mo
base plan$10.00
credits used15.00
included1,000
overage USD$0.05
// MODERATE USER
chat + 1 agentic / wk on Sonnet
$11.82/mo
base plan$10.00
credits used191.82
included1,000
overage USD$1.82
// HEAVY USER
daily Sonnet agentic + Opus brainstorms
$31.68/mo
base plan$10.00
credits used2,177.58
included1,000
overage USD$21.68

Run the CLI with your own usage shape and you get your number. The full report shows per-model breakdowns, per-workflow cost contribution, and the exact line item that's pushing you over.

What ships in v1.

WhatDoes what
meter.pyPython CLI. Runs locally. Three built-in profiles (light / moderate / heavy) plus a `--profile custom` mode that takes your own JSON shape. --json for piping into your team's tooling.
rates.json11 models from Anthropic, OpenAI, Google, xAI — all per-million token rates as published by GitHub for June 1. Human-readable. You can edit it yourself if rates shift.
meter_report.htmlSelf-contained HTML report. Inline SVG bar chart. No CDN, no JS framework. Opens in any browser, prints to PDF cleanly, sends to your manager without explaining anything.
Direct-API comparatorThe proprietary angle: for every workflow profile, computes what it would cost to call Anthropic / OpenAI / Google / xAI directly instead of through Copilot. Surfaces the crossover point in plain language.
README + rates_updateTwo short markdown files. Plain English. How to find your usage data inside GitHub Copilot, how to update rates.json yourself when GitHub publishes new rates.
CHANGELOGEvery rate-table revision logged. So when GitHub adjusts mid-month, you can replay your bill against the old rates.
Lifetime updatesFree updates to rates.json + the formula module forever. Delivered through the private GitHub repo invite you get on purchase.

The angle no one else will build.

Copilot vs direct API — the math GitHub will never show you.

Copilot's per-million token rates are published. The underlying providers' API rates are also published. Septim Meter is the only tool that does both calculations side-by-side and shows the crossover point.

For a heavy agentic user spending $21.78 of token value per month: direct API calls cost $21.78. Copilot adds $10 base for the IDE features. Question: is Copilot's tab-completion and IDE chrome worth $10/mo to you? For some buyers, yes. For others, no — but they need the math in front of them to decide.

GitHub will not ship this comparison. Their preview tool will show you what your Copilot bill looks like, not when leaving Copilot is cheaper. That's the gap Septim Meter fills.

Stack & delivery.

Python 3.9+ stdlib only — json, argparse, that's it. No pip install. No external dependencies. Runs identically on macOS, Windows, and Linux. Works on any solo dev machine. Suitable for production environments where audit-ability matters: every line is human-readable, no opaque library doing the math.

Delivered as a private GitHub repo invite. You get read access on purchase, the v1.0.0 ZIP is attached as a release asset, and lifetime free updates land in the same repo. No Septim-hosted infrastructure, no expiring download URLs, no SaaS dashboard to log into. You own the artifact.

Get notified when it launches.

Founding-rate access — first 50 buyers get FOUNDINGRATE24 (20% off, $23 instead of $29).

Septim Meter ships May 12 inside a 5-week window before GitHub's June 1 billing change goes live. Drop your email and I'll send you the buy link the morning it goes up. No drip sequence, no upsell — one email, one link.

No spam. One email at launch. Already on the list? You'll just get one. 14-day full refund post-launch, no questions asked.

Price.

$29
One-time, lifetime updates, no subscription
$29 once. Private GitHub repo invite, lifetime free updates to rates.json and the formula module, 14-day full refund. Built solo, runs solo, owned outright by you. Production-grade Python with stdlib-only dependencies. Suitable for enterprise environments where audit-ability matters.

Already burned by a runaway Anthropic bill? Septim Rescue ($299) is the emergency intervention — we get on a call within 4 hours and fix the circuit-breakers your pipeline is missing.

Want to read the math first? GitHub Copilot's June 1 token pricing — what it actually costs walks the formula and shows three worked examples.