Ghost Tab

Inspiration

We noticed a frustrating pattern while building with AI: our work was constantly getting fragmented.

We would research something in a browser, discuss it in ChatGPT, continue in Claude, inspect code on GitHub, read documentation, and then switch back to another AI — but none of these tools actually knew what had happened somewhere else.

Every new chat felt like starting over.

That led us to a simple question:

What if the context belonged to the user, instead of belonging to each AI tool?

That became GhostTab.

The Problem

AI assistants are powerful, but their context is trapped inside individual conversations and applications.

If you spend an hour researching a project in Claude and then open ChatGPT, you usually have to manually explain:

what you're building
what you've already tried
the decisions you've made
what sources you've read
what questions are still unanswered

As your work moves across websites and AI tools, valuable context gets lost.

The problem isn't that AI cannot remember.

The problem is that your work is fragmented across the web.

What GhostTab Solves

GhostTab creates a universal context layer for the browser.

It captures the pages and conversations you're working with, uses AI to understand what actually matters, and turns that information into structured workspace memory such as:

Goals · Decisions · Questions · Facts

Instead of making the user carry context from one AI to another, GhostTab carries the context for them.

The result is simple:

Your work follows you — not the AI.

How We Built It

GhostTab is built as a Chrome Manifest V3 extension using React, TypeScript, Vite, and Tailwind CSS.

We designed it around a local-first workspace architecture using IndexedDB, so workspace memory persists directly in the browser.

The pipeline works like this:

Browser → Context Capture → AI Understanding → Memory → Context Continuity

Our content scripts capture readable webpage content and, for supported AI platforms, extract conversation messages rather than just the currently visible text.

We then send relevant context to the Kilo AI Gateway, where the AI acts as a strict memory curator. Instead of saving everything, it identifies only durable information that can genuinely help continue the user's work.

GhostTab separates:

Sources — where information came from AI Insights — what the system understood Workspace Memory — what is worth remembering

Finally, GhostTab can generate a compact workspace context packet containing the user's goals, decisions, open questions, facts, and relevant sources. That context can be carried into another AI such as ChatGPT or Claude, allowing the user to continue without starting from zero.