ChatSpeed 2

Make long ChatGPT, Claude.ai, Grok, Gemini and DeepSeek conversations lighter and faster.

ChatSpeed 2 is an open-source Chromium browser extension designed to reduce browser rendering work in long AI conversations. It supports ChatGPT, Claude.ai, Grok, Gemini and DeepSeek with platform-specific optimizers.

Instead of deleting or modifying your conversation, ChatSpeed 2 adjusts eligible conversation requests so the browser receives a smaller recent portion of the chat.

Keep long AI conversations usable without deleting or storing your conversation content.

Built by Cobalt

---

Supported browsers

Browser	Status
Google Chrome	Supported
Brave	Supported and tested in normal use

ChatSpeed 2 is built as a Manifest V3 Chromium extension. Brave uses the same extension package; no separate Brave build is required.

---

Supported platforms

Platform	Optimization
ChatGPT	Request-level history reduction + render optimization
Claude.ai	Render optimization using CSS content-visibility and containment
Grok	Render optimization using CSS content-visibility and containment
Gemini	Render optimization using CSS content-visibility and containment
DeepSeek	Render optimization of mounted chat content without replacing DeepSeek's own virtualization

Claude, Grok, Gemini and DeepSeek support deliberately do not rewrite their private network requests. ChatSpeed accelerates long conversations on those platforms at the rendering layer and fails open if their page structure changes.

---

What ChatSpeed 2 does

Very large AI conversations can become increasingly heavy for the browser. More history can mean more messages to download, more text and code to process, more UI elements to render, and more browser memory usage.

ChatSpeed 2 reduces that workload before the page has to render it.

For example:

Normal request

ChatGPT
   ↓
Request 100 turns
   ↓
Browser receives 100 turns


ChatSpeed 2 — Balanced

ChatGPT
   ↓
ChatSpeed detects the request
   ↓
Request reduced to 16 turns
   ↓
Browser receives 16 turns

Your full conversation is not deleted. ChatSpeed only changes how much recent history is requested for an eligible page load.

---

How it works

ChatSpeed 2 runs a lightweight interceptor inside the ChatGPT page. It wraps the browser's normal fetch() function and watches for eligible ChatGPT conversation requests.

A request may look like:

/backend-api/conversations/<conversation-id>?num_turns=...

When ChatSpeed recognizes a safe request, it changes only the num_turns value and then lets the request continue through the browser normally.

ChatGPT UI
    │
    ▼
Conversation request
    │
    ▼
ChatSpeed interceptor
    │
    ├── Not eligible ─────► Original request
    │
    └── Eligible
           │
           ▼
       Reduce num_turns
           │
           ▼
       ChatGPT server
           │
           ▼
    Smaller response
           │
           ▼
     ChatGPT renderer

ChatSpeed does not replace ChatGPT's networking system. It makes a small request adjustment and then hands control back to the browser's original fetch() implementation.

---

Optimization modes

Mode	Target	Behavior
Safe	No reduction	Request optimization disabled
Balanced	16 turns	Good balance between context and performance
Turbo	8 turns	More aggressive optimization

Balanced is the default mode. The extension itself starts disabled until you enable it.

---

Safety behavior

ChatSpeed deliberately avoids modifying requests when it is not confident that optimization should be applied.

It does not rewrite:

• non-GET requests
• requests to another origin
• temporary chats
• pagination requests using before
• pagination requests using cursor
• requests without a valid num_turns
• requests already asking for fewer turns than the selected target
• requests while Safe mode is active

If something unexpected happens, ChatSpeed falls back to normal ChatGPT behavior.

Fail open. ChatGPT should continue working even if ChatSpeed cannot optimize a request.

---

Toolbar badge

ChatSpeed 2 includes a visible extension badge.

• Green ON means the extension is loaded and running.
• Flashing RUN appears briefly during detected ChatGPT page activity.

This gives you a quick visual confirmation that ChatSpeed is alive without opening Developer Tools.

---

Metrics

ChatSpeed keeps lightweight per-tab runtime metrics, including:

• rendered turns
• optimized requests
• avoided turns
• request optimizer status
• render optimizer status

Runtime metrics are stored in browser session storage. Settings are stored locally using Chrome extension storage.

---

Privacy

ChatSpeed 2 is designed to work locally inside your browser.

It does not require an external ChatSpeed server to optimize conversations.

The project is designed around these principles:

• no external ChatSpeed analytics
• no conversation database
• no remote ChatSpeed processing server
• no selling of conversation data

The optimization happens inside the active ChatGPT page.

---

Installation

ChatSpeed 2 currently installs as an unpacked Chromium extension.

1. Clone the repository

git clone https://github.com/echelong/ChatSpeed-2.git
cd ChatSpeed-2

2. Install dependencies

npm install

3. Build

npm run build

The extension will be created in:

dist/

4. Load in Chrome or Brave

Open the extensions page for your browser:

Chrome

chrome://extensions

Brave

brave://extensions

Then:

1. Enable Developer mode
2. Click Load unpacked
3. Select the dist folder
4. Pin ChatSpeed 2 to your toolbar

---

Project structure

src/
├── App.tsx
│   └── Extension popup
│
├── background.ts
│   ├── Settings
│   ├── Metrics
│   └── Toolbar badge
│
├── content.tsx
│   ├── Page bridge
│   └── Activity reporting
│
└── lib/
    ├── config.ts
    └── messages.ts

public/
└── chatspeed-interceptor.js
    └── ChatGPT request interceptor

---

Why optimize before rendering?

If the browser receives a huge conversation first, hiding old messages afterward does not remove all of the work that already happened.

ChatSpeed tries to reduce the workload earlier:

Request less
     ↓
Receive less
     ↓
Process less
     ↓
Render less

That is the core idea behind ChatSpeed 2.

---

Compatibility

Browsers

ChatSpeed 2 officially supports:

• Google Chrome
• Brave

Both use the same Manifest V3 build from the dist folder.

AI platforms

ChatSpeed 2 supports:

• ChatGPT
• Claude.ai
• Grok
• Gemini
• DeepSeek

ChatGPT uses both request-level and render optimization. Claude.ai, Grok, Gemini and DeepSeek use render-level optimization and do not receive ChatSpeed's ChatGPT request interceptor.

ChatSpeed depends on the current web structure of each supported platform. If a site changes its page structure or request format, ChatSpeed is designed to fail open rather than force an unsafe optimization.

---

Open source

ChatSpeed 2 is free to use under the MIT License.

You may:

• use it
• modify it
• fork it
• redistribute it
• build on it
• use it commercially

See LICENSE for the full terms.

---

Author

Built by Cobalt

GitHub: @echelong

---

Disclaimer

ChatSpeed 2 is an independent open-source project and is not affiliated with or endorsed by OpenAI.

ChatGPT is a product of OpenAI.