Unabyss gives Claude, ChatGPT and Cursor one shared memory built from Gmail, Slack and more via MCP. Plans from $13/mo with a 7-day free trial.
Researched and fact-checked by Bartu Cavusoglu · Updated October 26, 2026
Unabyss is a context layer for AI tools. It collects information from the apps a person already works in, including Slack, Gmail, Notion, Google Calendar and GitHub, and organizes it into one structured memory. Claude, ChatGPT, Cursor and any other MCP-compatible tool can then read that memory instead of starting each conversation from zero.
The problem it targets is familiar to anyone who uses more than one AI assistant. What one tool learns stays inside it, and the real details of a project often sit in email or chat where an AI cannot see them. Unabyss aims to close both gaps with a single source of truth that every connected agent shares.
Setup begins by connecting at least two sources, such as Gmail, Notion, Calendar, GitHub or Linear. Unabyss then extracts the content and structures it into a profile tagged by topic, source and sensitivity. The company describes this as a structured picture rather than a transcript.
The flow is not a one-time import. It runs in the background and works in both directions, so new emails and projects flow in while work done inside AI tools feeds back. Connected agents pull the relevant slice on demand over MCP, which means each tool receives only what is useful rather than a full history dump.
The core feature is automatic extraction and structuring of context from more than 25 integrated apps. Users can also edit files by hand, upload documents, chat with their context, or import existing context from ChatGPT custom instructions and Claude Projects. For tools without MCP support, 20+ one-click exports produce Markdown files that can be pasted anywhere.
Control is a second focus. Context is divided into Identity, Profile and Mind layers, and the Mind layer is opt-in only. Users approve, deny or edit access per layer and per file. Connections use scoped OAuth, access can be revoked instantly, and data is encrypted with AES-256 at rest and TLS 1.3 in transit.
Unabyss is a context layer for AI tools. It pulls information from apps like Slack, Gmail, Notion and GitHub into one structured memory that Claude, ChatGPT, Cursor and other MCP-compatible tools can read.
Pro is $13/month billed annually ($15 billed monthly), Max is $79/month billed annually ($89 monthly), and Team is $39 per seat per month billed annually ($49 monthly) with a 3-seat minimum.
Yes. Every plan starts with a 7-day free trial, and the Pro and Max trials do not require a credit card.
Built-in memories only cover conversations inside their own tool. Unabyss sits above all tools, pulls context from where work lives and serves it over MCP, so the same context follows the user across AI tools.
It connects to apps including Gmail, Slack, Notion, GitHub, Linear, Jira, Google Drive, Google Calendar and HubSpot, and works with MCP clients such as Claude, ChatGPT, Cursor, VS Code and Codex.
The company states that data is encrypted with AES-256 at rest and TLS 1.3 in transit, connections use scoped OAuth, and access can be revoked instantly. It says it does not sell data or train AI models on it. SOC 2 Type II is in progress.
MCP is the recommended method but not the only one. Users can also use one of 20+ one-click exports or a custom export, all delivered as Markdown files that can be pasted into any AI tool.
The site targets founders, builders, agencies and go-to-market teams who use several AI tools and keep their context in work apps.