Skip to content
MadanChaollaParkPublic

Repository files navigation

GPTandME icon

GPTandME

A simple, local prompt counter for your AI chats.

Chrome Web Store · Privacy · Support

GPTandME counts prompts you send on supported browser chat sites. See today's count at a glance, follow your week, and keep your history on your own device. No account or GPTandME server is required.

This repository contains the 1.6.0 source update. A local build is for testing; it does not update your installed extension or confirm publication to the Chrome Web Store.

What you get

  • A prominent daily count, plus this week, this month, and all-time totals.
  • A readable seven-day chart and today/all-time counts by service.
  • Today's count in the extension badge and an optional in-page counter.
  • Collapsed settings, backup/import tools, and diagnostics to keep everyday use focused.
  • Complete JSON backup/restore and CSV export/import for dated prompt counts.

Version 1.6.0 removes API cost estimates and thinking-time displays. It also removes the session, streak, last-24-hours, and model-breakdown panels. The updated page script no longer captures thinking time. Existing counts and legacy timing, model, hourly, and session data remain compatible with JSON backups; the storage and export schemas remain at v3. Reload already-open chat tabs after updating.

Supported sites

Service Hosts
ChatGPT chatgpt.com, chat.openai.com
Claude claude.ai
Gemini gemini.google.com
Perplexity perplexity.ai, www.perplexity.ai
Grok grok.com (optional access enabled in settings)

GPTandME counts new prompts sent after it is installed and active. Historical conversations are not backfilled. It does not count Claude Code, native desktop apps, command-line tools, direct API usage, or Chrome Incognito activity. It does not run on API, billing, or documentation sites.

On ChatGPT, Claude, Perplexity, and Grok, counting depends on a recognized outgoing user-send request. Request identifiers prevent duplicate counts. Gemini uses send activity from its primary page composer. A count records a submission, not a successful response: a request can count even if the provider later fails. Detection can miss sends if a provider changes its request endpoint or composer; those changes may require an extension update.

Install or try the update

Use the Chrome Web Store listing for the published extension. GPTandME 1.6.0 requires Chrome/Chromium 111 or newer.

To test the current source locally:

  1. Download or clone this repository.
  2. Open chrome://extensions and enable Developer mode.
  3. Select Load unpacked and choose the repository root.
  4. Reload every already-open supported chat tab.
  5. Send a new prompt and check that the badge and popup increment.
  6. For Grok, enable optional Grok counting in settings, allow access to grok.com, and reload existing Grok tabs.

Extension storage is tied to its extension ID. Make a JSON backup before changing installations; an unpacked copy may have a different ID and its own empty history.

Your data

Counts are stored by local date and service. Model labels, hourly counts, and bounded session records remain in the underlying data for compatibility and export, even though they are no longer shown as separate panels. Usage recorded before service attribution remains under unknown; the extension does not guess which service generated it. Week totals start on Monday, and month totals follow the calendar month.

A JSON backup preserves the complete normalized local dataset, including legacy thinking-time aggregates, model/hour/session data, settings, and diagnostics. Restoring a backup replaces your current data. Storage and export schema v3 remain unchanged.

CSV files contain prompt counts only, using date,provider,model,count. A CSV import adds counts to your current history; importing the same file twice duplicates them. Older date,model,count CSV files are still accepted and use provider unknown. CSV does not preserve timing, hourly, session, or settings data; use JSON to move a complete history.

Reset today removes today's dated counts and legacy timing aggregates and clears session aggregates and recent deduplication IDs. Older dated totals remain. Reset all removes all usage history. Both actions require confirmation in the popup.

Privacy

Usage data and preferences stay in chrome.storage.local. GPTandME does not store prompt text, response text, raw thinking labels, raw timing samples, or a page's full URL path. It has no analytics or external usage telemetry. You can manually download your data as a JSON backup or CSV file.

Existing permissions are unchanged from version 1.5.0. The required host permissions cover the two ChatGPT hosts used by the local request detector. Static scripts cover the other supported chat hosts; Grok scripts run only after optional access is granted. Request detection may inspect outgoing data locally to recognize a user send and its opaque event identifier, but it does not retain or transmit prompt text.

The extension keeps bounded local session and deduplication records to avoid duplicate counting. Legacy thinking-time aggregates remain in storage and backups; updated page scripts no longer collect new timing measurements. GPTandME cannot be enabled in Incognito mode.

Read the full privacy policy.

Development and release

Requires Node.js 20 or newer and the standard zip and unzip commands.

npm test
npm run build
npm run verify:package

The build creates dist/gptandme.zip with normalized archive metadata for reproducible bytes. Package verification checks the release contract, exact file inventory, and matching manifest/package versions.

The pinned GitHub Actions release workflow tests the tagged source, builds the ZIP, records its SHA-256 checksum, attests provenance, and publishes those exact files in a GitHub Release. Follow the release checklist before any Chrome Web Store submission. Documentation in docs/ is also source material; the public site is maintained in a separate repository.

Troubleshooting

  • A send is missing: confirm you are on a supported chat host, reload the extension, then reload the chat tab. Updates cannot replace content scripts already loaded in a page. If the provider changed its request endpoint or composer, report the affected service; detection may need an extension update.
  • A failed response still counted: totals record submitted prompts, not successful answers. A recognized outgoing request can count even if the response fails later.
  • The popup starts at zero: only new sends are counted. Check whether you loaded a second extension with a different ID; restore a JSON backup to move complete history.
  • Service totals do not cover the combined total: older unattributed usage remains unknown rather than being assigned to a service.
  • Grok stays at zero: enable its optional access in settings and reload any already-open Grok tabs.
  • A removed panel is missing: version 1.6.0 focuses on prompt counts. Existing legacy data remains in JSON backups.

For support, email [email protected] or use the public issue tracker. Include the extension/browser versions and affected service, without prompt text or conversation URLs.

GPTandME is independent and is not affiliated with, endorsed by, or sponsored by OpenAI, Anthropic, Google, Perplexity, or xAI.

Releases

Packages

Contributors

Languages