🔄 PromptSync PRO

Your context. Synced with AI.

Stop repeating yourself. PromptSync automatically injects your personal context into every AI chat — so you get smarter, more relevant answers without typing your preferences every time.

Works on: ChatGPT · Claude · Gemini · Perplexity · DeepSeek

Runs on: Chrome · Firefox · Edge · Brave · Opera · Zen · Firefox for Android

What is PromptSync?

Every new AI chat starts from zero. You find yourself typing the same thing over and over:

"I'm a developer. I use React and TypeScript. Keep answers short."

PromptSync fixes this.

It stores your profile and saved notes locally in your browser, then silently prepends them to every message before it sends. The AI sees your full context + your question as one rich prompt — giving you smarter, personalized responses instantly. And now, it works across all major browsers.

You type:      "How do I normalize a dataset?"

AI receives:   [USER CONTEXT START]
               Name: Ratul | Role: CSE Student, ML Learner
               Goals: Build real ML projects, master data science
               Memory: Uses Python, NumPy, Pandas — no code comments
               Instruction: Keep answer short. No extra explanation.
               [USER CONTEXT END]
               How do I normalize a dataset?

Pipeline: User Input → Parse Commands → Build Context → Inject → Track Tokens

Why Use PromptSync?

Problem	PromptSync Solution
Typing "I'm a developer, I use React..." every chat	Your profile auto-injects
AI gives long, unwanted explanations	Token-saving prompt enforces short answers
Forgetting to mention project context	Saved notes match to your question keywords
Different needs on different AI platforms	Per-platform profiles auto-switch
Wasting tokens on repeated context	Smart injection minimizes token usage

What's New in v3.1.1

🚀 Headline Feature: Cross-Browser Support

PromptSync PRO now runs natively on Chrome and Firefox — and by extension, all Chromium-based browsers (Edge, Brave, Opera) and Firefox for Android. One codebase. Two builds. Full parity.

Change	Details
🌐 Cross-Browser Support	Full feature parity on Chrome, Firefox, Edge, Brave, Opera, Firefox for Android
🔧 Unified Storage API	Abstracted `browser.storage` / `chrome.storage` layer — works on both engines with zero code duplication
🦊 Firefox Manifest	Separate `manifest.firefox.json` with `browser_specific_settings` for AMO submission
⏱️ Sandboxed Claude Counter	Fully independent token counter running only on `claude.ai`
🎭 Role-Based Engine	Strict `[ROLE]` / `[STYLE]` formatting with fail-safe detection
👁️ Preview System	Review and edit injected context before sending
📦 Dual Build System	`build.sh` now outputs `promptsync-chrome.zip` and `promptsync-firefox.zip` separately

Browser Support

Browser	Engine	Support	Notes
Chrome	Chromium	✅ Full	MV3 — Chrome Web Store ready
Firefox	Gecko	✅ Full	WebExtension API — AMO ready
Edge	Chromium	✅ Full	Load Chrome build directly
Brave	Chromium	✅ Full	Load Chrome build directly
Opera	Chromium	✅ Full	Load Chrome build directly
Zen Browser	Gecko	✅ Full	Load Firefox build — same as Firefox
Firefox for Android	Gecko	✅ Full	Install from Firefox Add-ons

Note: Safari is not currently supported. The floating panel's drag-and-drop relies on the Pointer Events API which has limited Safari compatibility.

Installation

Chrome / Edge / Brave / Opera

Step 1 — Build the Chrome package

git clone https://github.com/ratulhub/promptsync.git
cd promptsync
bash build.sh chrome
# Output: build/promptsync-chrome.zip  +  build/chrome/ (unpacked)

Step 2 — Load the extension

Open your browser and go to the extensions page:
- Chrome / Brave: chrome://extensions/
- Edge: edge://extensions/
- Opera: opera://extensions/
Enable Developer Mode (toggle in the top-right corner)
Click Load unpacked → select the build/chrome/ folder
Pin the PromptSync icon to your toolbar ✅

Requires Chrome 88+ / Edge 88+ / Brave 1.20+ (Manifest V3).

output.mp4

📦 Installation walkthrough — follow along to get set up in under a minute.

Firefox / Zen Browser / Firefox for Android

Step 1 — Build the Firefox package

git clone https://github.com/ratulhub/promptsync.git
cd promptsync
bash build.sh firefox
# Output: build/promptsync-firefox.zip  +  build/firefox/ (unpacked)

Step 2A — Temporary Load (about:debugging) ✅ Recommended for local testing

Open Firefox / Zen → navigate to about:debugging
Click This Firefox in the left sidebar
Click Load Temporary Add-on...
Navigate to build/firefox/ → select manifest.json ✅

⚠️ Do NOT use about:addons → "Install Add-on From File" for local testing. Firefox will flag the extension as "corrupt" because it is unsigned by Mozilla. Always use about:debugging for unpacked local loads.

⚠️ Temporary add-ons are removed when Firefox restarts. Use Step 2B for persistence.

Step 2B — Signed Installation (Persistent)

Open Firefox → go to about:addons
Click the ⚙️ gear icon → Install Add-on From File...
Select build/promptsync-firefox.zip ✅

The extension must be self-signed or published on addons.mozilla.org (AMO) for full persistence without Firefox Developer Edition.

Firefox for Android

Install Firefox for Android from the Google Play Store
Enable custom add-ons: Firefox menu → Settings → About Firefox → tap the logo 5× to unlock debug menu
Go to Settings → Install Extension from File → select promptsync-firefox.zip
Or install directly from the AMO listing once published ✅

Quick Start

Click the PromptSync icon → open the popup
Go to Profile tab → fill in your name, role, and goals → Save Profile
Go to Saved tab → add context notes (e.g. "I use Python for all projects")
Open any supported AI site (ChatGPT, Claude, Gemini...)
Type a message and send — your context injects automatically ✅

You'll see a token badge (e.g. 85 tok) on the page confirming it's active.

Features

1. Multi-Profile System

Store multiple profiles for different roles and switch between them instantly.

Built-in Profiles	What It Stores
Default	General purpose
Student	Study-focused context
Developer	Coding-focused context
Custom	Anything you define

Each profile holds: Name, Role, Goals, Skills, Preferences, Custom Notes

Tip: Create profiles for Work, Study, Personal — switch from the popup dropdown in one click.

2. Role-Based Engine

Choose a professional role so the AI responds exactly how you need. Uses strict [ROLE] and [STYLE] formatting for reliable results.

Role	AI Behavior
🧑‍💻 Developer	Clean minimal code, no unnecessary explanation
🔬 Researcher	Thorough analysis, structured and sourced
🎨 Designer	UI/UX focused, visual thinking
✍️ Writer	Clear, structured, professional writing
💼 Business	Formal, professional business-style answers

You can also create your own custom roles from the Roles tab.

Fail-Safe Detection: If you type "Act as..." or "You are a..." manually, PromptSync detects it and skips injecting its own role — so it never conflicts with your intent.

3. Smart Context Injection

Builds a context block using your profile + selected notes before every message.

Default template:

[USER CONTEXT START]
Name: {user_name}
Role: {user_role}
Goals: {user_goals}
Memory: {memories}
Instruction: {instruction}
[USER CONTEXT END]

Each section can be toggled on/off individually
Template is fully editable from Settings → Template
Notes are selected smartly based on keyword relevance to your message

4. Slash Commands

Type these directly in the chat input to control injection for that one message. Commands are removed before sending — the AI never sees them.

Command	What It Does
`/no-memory`	Skip all injection for this message
`/only-profile`	Inject profile only, skip saved notes
`/strict`	Inject name + pinned notes only
`/code`	Switch to coding mode
`/short`	Force the shortest possible answer
`/explain`	Allow full detailed explanation
`/detail`	Full context + detailed response
`/mode <name>`	Activate a named task mode
`/temp <text>`	Add session-only memory (not saved)
`/reset`	Revert to default injection behavior

Example:

/code How do I debounce a React input?

5. Custom Commands

Create your own slash commands with predefined behaviors.

Go to Modes tab → Custom Commands → + New Command
Set a name (e.g. /fast) and define what it does:
- Switch to a task mode
- Enable/disable memory injection
- Inject a custom instruction

No arbitrary code execution — only safe predefined actions.

6. Task Modes

Replace the default instruction with a specialized one for different types of work.

Mode	Injected Instruction
Coding	"Give clean, minimal, working code. No explanation unless asked."
Study	"Explain simply step by step. Use easy words. Give examples."
Short	"Answer in the shortest possible way."
Custom	Your own instruction text

Set globally from popup or floating panel
Override per-message with /mode <name>
Disable for one message with /reset

7. Saved Context Notes

A full note management system for persistent context.

Feature	Detail
Text	The context content to inject
Tags	Labels for organization and filtering
Priority	High / Medium / Low — affects injection scoring
Pin	Always inject regardless of relevance
Enable/Disable	Toggle without deleting
Search	Filter by text or tag
Drag & Drop	Reorder notes visually
Bulk Delete	Select multiple and remove at once

How notes are selected: Keywords are extracted from your message → each note is scored by keyword match + priority + pin bonus → top N notes are injected.

Right-click any text on a webpage → "Save to PromptSync" to instantly create a note from selected content.

8. Preview System

Want to see exactly what's being injected before it sends?

Open popup → Roles tab → enable Preview before sending
Press Enter in the chat — PromptSync injects the context into the input box and pauses
Review or edit the text
Press Enter again to actually send

9. Sandboxed Claude Counter

A fully isolated, independent token counter running exclusively on claude.ai — never interfering with the core injection pipeline.

Feature	Detail
Real-time BPE tokenization	Tracks input, memory, and session tokens live
Context window progress bar	Visual indicator for Claude's 200K window
Glassmorphism floating UI	Draggable panel integrated into the Claude interface
Zero conflicts	Sandboxed execution — completely independent from the core engine

10. Floating Panel & Token Tracking

A persistent overlay on all supported AI pages.

Token Badge — always visible, shows current injection token count (e.g. 120 tok). Click to expand the full panel.

Full Panel controls:

Memory injection toggle (ON/OFF)
Mode switcher dropdown
Token breakdown: input / memory / session
Context window progress bar (green → yellow → red)
Session reset button

Context window limits per platform:

Platform	Limit
ChatGPT	128,000 tokens
Claude	200,000 tokens
Gemini	1,000,000 tokens
Perplexity	16,000 tokens
DeepSeek	64,000 tokens

Token counting uses the o200k_base BPE tokenizer (same as GPT-4o) for accuracy. The panel is draggable — position saved across sessions.

11. Memory Decay

Prevents your saved notes from becoming stale and bloated.

Every note tracks a lastUsed timestamp
After a configurable number of days (default: 7 days), unused notes are flagged as stale
A decay banner appears in the Saved tab → click "Remove All Stale" to clean up
Pinned notes are exempt — they never decay

Configure in Settings → Decay.

12. Security Guard

Scans note text for sensitive data before saving.

Detected Pattern	Example
OpenAI API keys	`sk-...`
Stripe keys	`pk_live_...`
AWS access keys	`AKIA...`
GitHub tokens	`ghp_...`
Private keys	`-----BEGIN PRIVATE KEY-----`
Bearer tokens	`Bearer ...`
Passwords	`password:`, `secret:`
Credit cards	Luhn-validated 13–19 digit numbers

Shows a warning with override option. Toggle in Settings → Security.

13. Profile Auto-Switch

Automatically load the right profile when visiting each AI platform.

Example mappings:

claude.ai → Student
chatgpt.com → Developer
gemini.google.com → Default

Configure in Settings → Auto-Switch. Toggle on/off without losing your mappings.

14. Customizable Template

Edit the exact format of the injected context block.

Use placeholders: {user_name}, {user_role}, {user_goals}, {memories}, {instruction}
Toggle each section on/off individually
Changes apply to all future injections immediately

15. Import & Export

Export: Settings → Advanced → Export Data → saves promptsync-*.json
Import: Settings → Advanced → Import Data → select .json file

Exports include all profiles, notes, modes, commands, and settings. Input is validated on import.

16. Dark Mode

Light / Dark / System (auto OS). Cycle with the button in the popup header. Glassmorphism effects on the floating panel and toast notifications.

The Popup Dashboard

The popup has 5 tabs:

Tab	What's Inside
Profile	Profile switcher, form fields, injection preview, token bar
Saved	Add/manage notes with tags, priority, pin, decay banner
Modes	Task modes + custom commands
Settings	Template, response control, security, decay, auto-switch, platforms, advanced
Help	All commands listed, tips, reset onboarding

Keyboard Shortcuts

Shortcut	Action
`Ctrl + Shift + M`	Open popup (Windows / Linux)
`Cmd + Shift + M`	Open popup (Mac)
Right-click selected text	"Save to PromptSync" — instantly save as a note

Architecture

promptsync/
├── manifest.json                  — Chrome / Chromium (Manifest V3)
├── manifest.firefox.json          — Firefox (WebExtension + browser_specific_settings)
├── build.sh                       — Dual build: promptsync-chrome.zip + promptsync-firefox.zip
├── background/
│   └── background.js              — Service worker (context menu, shortcuts, alarms)
├── claude-counter/                — Sandboxed Claude Token Counter (independent)
│   ├── content/
│   │   ├── bridge-client.js
│   │   ├── constants.js
│   │   ├── main.js
│   │   ├── tokens.js
│   │   └── ui.js
│   ├── injected/
│   │   └── bridge.js
│   ├── vendor/
│   │   └── o200k_base.js
│   └── styles.css
├── content/
│   ├── content.js                 — Core injection pipeline
│   ├── content.css                — All injected UI styles
│   ├── floatingPanel.js           — Token badge + control panel
│   └── onboarding.js              — First-run welcome overlay
├── popup/
│   ├── popup.html                 — 5-tab dashboard
│   ├── popup.js                   — All popup logic
│   └── popup.css                  — Styles with CSS variables
├── storage/
│   └── storage.js                 — Unified API: chrome.storage / browser.storage abstraction
├── utils/
│   ├── tokenizer.js               — BPE token counting (o200k_base)
│   ├── contextEngine.js           — Smart context selection + template engine
│   ├── commandParser.js           — Slash command parser + custom commands
│   ├── tokenTracker.js            — Token tracking integration
│   └── securityGuard.js           — Sensitive data detection
├── vendor/
│   └── o200k_base.js              — GPT tokenizer library (offline, bundled)
└── icons/
    ├── icon16.png
    ├── icon48.png
    └── icon128.png

Data Flow:

User Input → Parse Commands → Build Context (Engine) → Inject → Track Tokens
                ↓                       ↓
          Custom Commands     Template + Profile + Notes + Mode

Cross-Browser Storage Abstraction (storage/storage.js):

// Works on Chrome (chrome.*) and Firefox (browser.*)
const api = typeof browser !== "undefined" ? browser : chrome;
api.storage.local.get(keys, callback);

Storage keys (all local — never synced to a server):

Key	Stores
`asm_profiles`	All profile data
`asm_items`	Saved context notes
`asm_settings`	All settings
`asm_task_modes`	Built-in + custom modes
`asm_roles`	Role-Based Engine data
`asm_custom_commands`	User-defined slash commands
`asm_template`	Injection template format
`asm_panel_position`	Floating panel coordinates

Privacy & Security

100% local — all data stored in chrome.storage.local / browser.storage.local. Nothing ever leaves your device.
Zero network requests — no analytics, no telemetry, no external API calls.
XSS prevention — all output is HTML-escaped in display contexts.
Security Guard — scans for API keys, passwords, and credit cards before every save.
No code execution — custom commands use predefined action sets only. No eval.
Namespace isolation — all DOM elements prefixed with asm- to avoid conflicts.
Open source — inspect every line of code.

Changelog

v3.1.1 (Latest)

🌐 Cross-Browser Support — full parity on Chrome, Firefox, Edge, Brave, Opera, Firefox for Android
🔧 Unified Storage API — single browser/chrome abstraction layer, zero code duplication
📦 Dual Build System — build.sh outputs promptsync-chrome.zip and promptsync-firefox.zip
🦊 Firefox Manifest — dedicated manifest.firefox.json with browser_specific_settings for AMO
⏱️ Sandboxed Claude Counter — dedicated isolated token counter for Claude
🎭 Role-Based Engine — strict [ROLE] / [STYLE] formatting with fail-safe detection
👁️ Preview System — review and edit injected context before sending
🛡️ Fail-Safe Detection — skips role injection if you manually define a role in your message

v3.0.0

Multi-profile system with profile auto-switch per domain
Floating panel with real-time BPE token tracking
Memory decay system with configurable threshold
Security Guard with Luhn credit card validation
Custom commands + customizable task modes
Editable injection template with per-section toggles
Dark / Light / System theme with glassmorphism
Onboarding 3-step overlay on first install

v2.5.0

Token tracking, command system, context engine, dark mode

v2.0.0

Smart injection, keyword matching, platform support

v1.0.0

Basic memory storage and injection

Author

MD. Abdur Rahim Ratul

License

PromptSync PRO v3.1.1 — Stop repeating. Start syncing.

Built by MD. Abdur Rahim Ratul

_{Claude token counter inspired by she-llac/claude-counter}

Generated By MD. Abdur Rahim Ratul

Name		Name	Last commit message	Last commit date
Latest commit History 50 Commits
background		background
build-chrome		build-chrome
build-firefox		build-firefox
build		build
claude-counter		claude-counter
content		content
counter_ref		counter_ref
icons		icons
popup		popup
storage		storage
utils		utils
vendor		vendor
.gitignore		.gitignore
FEATURES.md		FEATURES.md
LICENSE		LICENSE
Manual.txt		Manual.txt
README.md		README.md
USER_MANUAL.md		USER_MANUAL.md
browserApi.js		browserApi.js
build.sh		build.sh
manifest.chrome.json		manifest.chrome.json
manifest.firefox.json		manifest.firefox.json
manifest.json		manifest.json
overview.txt		overview.txt

Folders and files

Latest commit

History

Repository files navigation

🔄 PromptSync PRO

Table of Contents

What is PromptSync?

Why Use PromptSync?

What's New in v3.1.1

Browser Support

Installation

Chrome / Edge / Brave / Opera

Step 1 — Build the Chrome package

Step 2 — Load the extension

Firefox / Zen Browser / Firefox for Android

Step 1 — Build the Firefox package

Step 2A — Temporary Load (about:debugging) ✅ Recommended for local testing

Step 2B — Signed Installation (Persistent)

Firefox for Android

Quick Start

Features

1. Multi-Profile System

2. Role-Based Engine

3. Smart Context Injection

4. Slash Commands

5. Custom Commands

6. Task Modes

7. Saved Context Notes

8. Preview System

9. Sandboxed Claude Counter

10. Floating Panel & Token Tracking

11. Memory Decay

12. Security Guard

13. Profile Auto-Switch

14. Customizable Template

15. Import & Export

16. Dark Mode

The Popup Dashboard

Keyboard Shortcuts

Architecture

Privacy & Security

Changelog

v3.1.1 (Latest)

v3.0.0

v2.5.0

v2.0.0

v1.0.0

Author

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 1

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages