Yo GitHub Copilot: are we cooked?

Estimate your GitHub Copilot AI-credit cost in preparation for June 1st. Pulls usage from VS Code, OpenCode, Pi, Zed, GitHub Copilot CLI, and Copilot for Xcode, aggregates it based on the new per-token pricing, and generates a local HTML report. Fully local.

Relevant links

• April 27th, 2026: GitHub Copilot is moving to usage-based billing
• Models and per-token pricing for GitHub Copilot
• Usage-based billing for individuals
• Usage-based billing for organizations and enterprises

TL;DR: GitHub Copilot is moving from premium-request quotas to per-token billing on June 1st. Agentic workflows can now cost more than the old premium-request mental model suggests.

Just let your agent do it

Don't want to run this yourself? Paste this prompt into your coding agent:

Clone https://github.com/PanAchy/copilot-arewecooked, install dependencies, build it, and run it.
Then open the generated HTML report and tell me whether or not I'm going to be cooked under the new Copilot AI-credit billing.

Setup

The quickest way (requires Node.js 20+):

npx copilot-arewecooked

Or clone and build:

git clone https://github.com/PanAchy/copilot-arewecooked.git
cd copilot-arewecooked
npm install
npm run build
npm run generate

By default, this writes an HTML report and PNG screenshot like:

copilot-report-YYYY-MM-DD-abc123.html
copilot-report-YYYY-MM-DD-abc123.png

Options

| Flag | Description | | ---------------------- | ----------------------------------------------------------------------------------------------- | | --days <n> | Days to look back (default: all available data) | | --since <date> | Only include records from this date onward (YYYY-MM-DD) | | --auto-model <model> | Treat records reported as auto as a specific model for cost estimation (e.g. gpt-5.3-codex) | | --json | Print detailed normalized JSON to stdout | | --html [path] | Write HTML report to a specific path |

npm run generate -- --days 30
npm run generate -- --since 2026-04-15
npm run generate -- --auto-model gpt-5.3-codex
npm run generate -- --html report.html
npm run generate -- --json

Note on auto model: Some tools (e.g. VS Code Copilot) may report the model as auto when the user has not selected a specific model. Credits for those requests will show as zero unless you specify --auto-model to map them to a known priced model. The HTML report will include a note when any auto records are remapped.

How data is extracted

| Source | Paths | Token accuracy | | --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------- | | VS Code | ~/Library/Application Support/Code{, - Insiders}/User/workspaceStorage/*/chatSessions/*.jsonl (macOS) · %APPDATA%/Code{, - Insiders}/User/workspaceStorage/*/chatSessions/*.jsonl (Windows) · ~/.config/Code{, - Insiders}/User/workspaceStorage/*/chatSessions/*.jsonl (Linux) | Input estimated, output exact, cache not persisted | | OpenCode | ~/.local/share/opencode/opencode.db (macOS and Linux) · %LOCALAPPDATA%/opencode/opencode.db / %APPDATA%/opencode/opencode.db (Windows) | All exact (input, output, cache read/write) | | Pi | ~/.pi/agent/sessions/**/*.jsonl (all platforms) | All exact (input, output, cache read/write) | | Zed | ~/.local/share/zed/threads/threads.db (falls back to thread.db if present) | All exact (input, output, cache read/write) | | Copilot CLI | ~/.copilot/session-state/*/events.jsonl (all platforms) | Output exact, input estimated, compaction exact | | Copilot for Xcode | ~/Library/Logs/GitHubCopilot/*.log (macOS only) | All exact (input, output, cache read); model attribution via heuristic |