# 🧠 Memorall — Unikorn submission profile

## 🏷️ Product name

Memorall

## 💬 A positioning phrase (tagline)

Memorall turns your browser into a workspace for AI agents: with long-term memory, real tools, and you choose the model — running natively on your machine.

## 📝 Product description (paste into description box)

🧠 Long-term memory · 🛠️ Real tools · 🎛️ Choose your model · 🔒 Local-first

### ✨ What the product is

Memorall is a browser-native agent workspace — an extension for Chrome/Edge (including a web version and a desktop version sharing the same codebase) that transforms your browser into a full workplace for AI agents:

- 🧠 Long-term memory in the form of a knowledge graph by topic

- 📚 Document library that can read PDF, images, Excel right in the app

- 🛠️ Real manipulation tools: browser, files, sandbox to run code

- 🎛️ Model selection rights: run locally or call APIs when necessary

All data is stored locally on the user's machine, login is not required, and data is not pushed to the cloud.

### 🎯 Problems solved

Most knowledge work occurs in browser tabs, but current AI tools are disconnected from there:

- 🗂️ You read documents, open dozens of tabs, save PDFs, take notes…

- 📋 …and still have to manually copy–paste each segment into a chatbot

- 🔁 …and every conversation starts from scratch, as the assistant remembers nothing about your project

- ⚠️ Accompanied by privacy risks: internal documents, contracts, customer data being sent to third-party servers

Memorall solves both. What you read and actively save — web pages, highlighted segments, screenshots, PDFs, Excel sheets, notes — are transformed into structured memories: a topic-based knowledge graph, combining keyword searches and semantic searches using vectors. The agent understands the context of your project, terminology, and your long-term research trails instead of just responding to disconnected prompts. And since everything runs locally, the data remains on your machine.

### 🚀 Key features

1. 💬 Assistant right on the page. Open a chat panel on any website; attach highlights, displayed content, entire text/HTML, or screenshots for context; save to a topic with just a few clicks.

2. 🕸️ Memory and knowledge graph. Convert text, Markdown, PDF, Excel into nodes and relationships by topic; browse visually with graphs, perform semantic searches, edit and clean data right in the interface.

3. 🦾 Agent with real tools, not just text responses. Read/write document library and workspace files, control the browser (open pages, read DOM, wait for selectors, manipulate), and run Node.js code in a sandbox right within the browser — install packages, run servers, build test applications.

4. 🧩 Create your own agent. Each agent is a graph defined by you: system prompts, skills, tools, connect to MCP (HTTP/SSE), set scheduled tasks; enable/disable each capability (knowledge retrieval, web browser, sandbox, file system, delegation) for each agent.

5. 🎛️ Freedom to choose models. Run models directly in the browser with WebLLM, Wllama, or Transformers.js (accelerated with WebGPU when supported), or connect to OpenAI, OpenRouter, Ollama, LM Studio for more powerful models. Choose embeddings of 384d / 768d / 1536d.

🔐 Required extension permissions and reasons: activeTab + content scripts (read page content upon your request) · storage (save configurations and local data) · tabswebNavigation (open pages and track navigation for agents) · contextMenus (quickly save highlighted content) · notifications