Unabyss

Unabyss is an easy way to deliver the AI context it needs to work correctly. It pulls your information from tools you use daily - Slack, Gmail, Notion, GitHub, meetings - into one structured memory, so Claude, Cursor, ChatGPT, or any MCP-compatible tool can read it. You stop re-explaining yourself every time you switch tools.

Right now, every chat with AI starts from zero - you re-explain your role, your project, your preferences, every time. What ChatGPT learns stays trapped in ChatGPT. Connect your sources once, and any AI tool that supports MCP already knows your context before you type a word. Never re-explain yourself to an AI again.

Under 90 seconds. Connect the sources you want to start with - Unabyss extracts your context stored there. Then connect Unabyss to Claude Desktop or Cursor via MCP, or export your context files directly and paste them into any AI tool.

You connect the sources you want to start with. Available at launch: LinkedIn, X, your personal website, Gmail, Notion, Slack, Google Docs, Google Calendar, Google Drive, GitHub, and Obsidian. You can also import existing context from ChatGPT custom instructions and Claude Projects -- so context you've already built elsewhere isn't lost.

ChatGPT Memory and Claude Memory each remember conversations inside that one tool - the memory doesn't travel with you. Unabyss sits above all your tools: it pulls context from where your work actually lives and serves it via MCP. Switch from Claude to Cursor to ChatGPT and the context comes with you, not just what you told one chatbot.

A context file works - until it goes stale. It's a snapshot: your role, your project, your priorities change, and the file doesn't. You're stuck manually updating and re-pasting it into every tool. Unabyss stays connected to your actual sources, so it updates itself - and it also pulls from what you do inside Claude, ChatGPT, or Cursor, feeding that back in. A .md file can't do that.

AI memory is what a platform learns about you from past conversations. It's reactive, unstructured, and locked to the platform that built it -- switch tools and you start from zero. AI context is intentional and pre-extracted: who you are, structured from authoritative sources, ready to serve to any tool before the first message. Memory is a side effect of using AI. Context is infrastructure you own.

A personal context layer is a structured, portable store of your identity, professional background, and preferences that any AI tool can access -- with your permission -- before the first interaction begins. Unlike AI memory (which is built reactively from past conversations and locked to one platform), a context layer is intentional and pre-extracted from authoritative sources like LinkedIn, Notion, or Gmail. Unabyss is the context layer: the place where your context lives, independent of any single AI tool.

You don't have to. Unabyss stays connected to your sources and updates your context vault automatically as your situation changes -- new role, new project, new priorities. You can also update any file manually at any time, upload documents, or chat with your context directly to add things the extraction didn't capture.

MCP (Model Context Protocol) is an open standard, co-developed by Anthropic, OpenAI, and Block, that lets AI tools pull structured data from external sources. When connected via MCP, Claude, Cursor, or any MCP-compatible agent pulls your context automatically at the start of each session. MCP is the recommended way to use Unabyss, but not the only way. You can also use one of 20+ one-click exports -- pre-formatted context files ready to paste into a specific tool -- or create a custom export. All exports are Markdown files you can upload anywhere.