Product Foundation
Product Identity
- Product family: Invariant
- Product name: Invariant LLM Connect
- Parent company: Cat vs Dot Technologies
- Product type: dedicated lightweight browser for web-based LLM platforms
- Primary descriptor: A lightweight browser for LLM platforms.
- Core message: Give the resources to the model, not the browser.
The product name is written exactly as Invariant LLM Connect. Do not shorten it publicly to Invariant Connect, Invariant Browser, Invariant AI Connect, LLM Connect Browser, or Invariant LLM Browser.
Definition
Invariant LLM Connect is a purpose-built browser whose sole function is to run web-based LLM platforms with substantially less unnecessary system overhead than a conventional general-purpose browser.
It is intentionally narrow. It should coexist with the user's daily browser and should not evolve into a normal browser with AI branding.
In Scope
Invariant LLM Connect:
- loads existing web-based LLM interfaces rather than recreating them
- supports local, LAN, VPN, and remote LLM-platform URLs
- supports multiple LLM platforms
- supports multiple tabs and windows
- allows several platforms or sessions to run side by side
- hands external websites to the user's normal system browser
- minimizes browser-controlled RAM, CPU, GPU, VRAM, disk, and network overhead
- preserves required modern web security functionality
- remains platform-agnostic rather than tied to Open WebUI or any other single product
Open WebUI, AnythingLLM, LibreChat, and similar applications are examples of software Connect can host. None defines the product itself.
Out of Scope
Connect is not:
- a replacement for Firefox, Chrome, Edge, Safari, or another daily browser
- an LLM frontend
- an inference server
- a model manager
- an agent framework
- a RAG system
- an AI assistant
- a search engine
- a general-purpose web browser
- an Open WebUI client specifically
- a Chromium wrapper
- an Electron application
The loaded LLM platform remains responsible for its own chat interface, model controls, agents, RAG, tools, images, files, settings, authentication, and other capabilities.
Target Users
The primary audience is anyone using web-based LLM platforms who does not want a general-purpose browser consuming resources alongside inference.
Especially relevant users include local-LLM users, low-resource computers, older laptops, mini PCs, AI workstations, home labs, dedicated AI terminals, machines near RAM or VRAM limits, and systems where the LLM server runs on another device.
High-end systems can also benefit when the user wants LLM activity separated from normal browser sessions.
Product Philosophy
Every proposed feature should answer:
Does this directly improve the user's ability to operate a web-based LLM platform?
If not, it should normally be rejected.
Resource efficiency is a functional requirement, not a marketing adjective. Security remains mandatory; a lightweight application that becomes unsafe is not successful.
Fixed Initial Foundation
- Name: Invariant LLM Connect
- Family: Invariant
- Purpose: lightweight browser exclusively for LLM web platforms
- Preferred engine: Gecko
- Fallback engine: none; the product runtime is Gecko-only
- General web browsing: no
- Normal address bar: no
- User-defined LLM endpoints: yes
- Known-platform presets: optional
- Tabs: yes
- Multiple windows: yes
- External sites: system browser
- Existing LLM frontend rendered directly: yes
- Native replacement LLM interface: no
- Resource efficiency: core product requirement
- Unused inherited browser functionality: remove where practical
- Security systems: preserve
- Personal/private AI references in public product: none
Changes that materially alter this boundary are design decisions, not incidental implementation details.