How I built thredly: a Chrome extension that never lets your AI context window die

thredly — Thu, 04 Jun 2026 20:31:43 +0000

If you use ChatGPT, Claude, or DeepSeek heavily, you've hit this wall: the context window fills up, responses get worse, and you're forced to start a fresh chat — losing everything you built up.

I kept running into this. So I built thredly — a Chrome extension that summarises your entire AI conversation and carries the context into a new thread automatically.

The problem

Long AI sessions are where the real work happens. You've refined your prompts, established shared context, got the model thinking in a specific way. Then it degrades. You either push through with broken context or start over and lose momentum.

What thredly does

It reads your current chat, generates a structured summary — goals, decisions, key context — and opens a new thread pre-loaded with it. The AI picks up exactly where you left off.

Works on ChatGPT, Claude, and DeepSeek. Takes about 5 seconds.

The build

It's a Chrome extension built with vanilla JS. The main challenge was reliably extracting conversation structure across three different AI interfaces — each renders their chat DOM completely differently.

For the summary itself, I'm calling the OpenAI API with a prompt engineered to preserve what actually matters: intent, decisions made, open questions, and established constraints.

Where it stands

235+ installs
5.0 star rating
Paying customers on Pro and Ultimate tiers

Still early days. If you've felt this pain, give it a try: thredly.io

Happy to answer any questions on the build in the comments.

How to continue long ChatGPT and Claude chats without losing context

thredly — Mon, 27 Apr 2026 19:38:45 +0000

Long AI chats are useful until the thread becomes too heavy to continue.

At the start, the model understands the project. It remembers your constraints, your decisions, your preferred structure, and the direction of the work. But after enough back-and-forth, the quality often starts to drift.

The problem is not just that the chat is long.

The problem is that the useful context is trapped inside that long thread.

When you start a new chat, you get speed and clarity back, but you usually lose the working context that made the old conversation valuable. That means repeating your goals, explaining decisions again, and trying to rebuild the thread manually.

A better approach is to carry the important context forward.

That is what I built thredly for.

thredly helps people continue long ChatGPT, Claude, and DeepSeek conversations without losing context. It creates a clean handover from the old thread so you can paste it into a new chat and carry on from where you left off.

Try it here: