MCP App Store
by KeenEthics
Sabi icon

Sabi MCP Server for ChatGPT

By Inovo Studios, LLC

Overview

Sabi is an AI assistant you text. Over time it learns who you are, the people you work with, how you like things done and what you are working on, and it keeps a Library of documents you have saved. Connect Sabi to ChatGPT and that knowledge comes with you. ChatGPT can: - Recall what Sabi knows about you, your people, your preferences and your projects - Search and read the documents in your Sabi Library - Remember a new fact you want Sabi to keep - Save a note or summary to your Library - Forget a memory you no longer want kept You connect by signing in to Sabi and approving access. No API key. You can disconnect at any time from Connections in Sabi. ChatGPT only sees your own Sabi account. A Sabi account is required. Start at meetsabi.com.

How to connect

  1. ChatGPT: Open the Sabi listing in the ChatGPT app directory and choose Connect. ChatGPT listing Sabi exposes 7 tools here.

Tools

sabi_forget

ChatGPT
Delete one stored fact, when the user says it is wrong, out of date, or asks you to forget it. Requires the id of a memory returned by sabi_recall_memories in this conversation — never guess an id. This is also how you correct a fact: saving the new version with sabi_remember does not retire the old one, so do both — forget the stale fact, then remember the new one. Leaving the old one in place means Sabi holds two contradicting facts and may recall either until it reconciles them overnight.

sabi_list_documents

ChatGPT
List what is in the user's Sabi Library, newest first, including documents still being indexed. Images carry a note describing what the picture shows ("house floor plan cutaway") — match the user's words against those notes, because filenames are often cryptic. Use this to find a document id, or to answer "what do I have saved about X".

sabi_read_document

ChatGPT
Read a whole Library document, by id from sabi_search_library or sabi_list_documents. Long documents come back a page at a time: when the result says has_more, call again with start_ordinal set to next_ordinal to continue. If you only need one fact from a long transcript, sabi_search_library is cheaper and usually enough.

sabi_recall_memories

ChatGPT
Search what Sabi knows about this user — the durable facts they have built up over time: their people, their preferences, their commitments, decisions they have made, how they like things done. Call this before answering anything personal about them, their work, or the people around them, rather than asking them to repeat it. Returns the most relevant stored facts, newest and strongest first.

sabi_remember

ChatGPT
Save a durable fact about the user to Sabi, so every agent they connect knows it from now on — including Sabi itself when they text it. Use this when they tell you something about themselves, their people, or how they want things done that would be worth knowing next month. Do not save passing context, anything they asked you to keep to this conversation, or something you inferred rather than were told. One fact per call, stated plainly in the third person ('Bruce's cofounder is Anna'). Facts longer than 500 characters are trimmed, so keep it to the fact itself. This only adds: if the new fact replaces one you found with sabi_recall_memories, delete that one with sabi_forget, or Sabi will hold both.

sabi_save_to_library

ChatGPT
Save a piece of text to the user's Sabi Library so it is searchable later from any agent they connect, and by Sabi itself — research you have pulled together, a summary they want to keep, notes from this conversation. Ask first unless they asked you to save it. Every call saves a new document, even when the title matches one already there, so don't call it twice for the same note — there is no way to edit a saved note from here. Saving is not instant: indexing runs after this returns, so a search for it may come up empty for a minute.

sabi_search_library

ChatGPT
Search the user's Sabi Library — documents, notes, meeting transcripts and images they have saved — and get back the passages that match. This is the right first call for anything that would be written down rather than remembered: a contract, a transcript, a spec, notes from a meeting. Returns excerpts with the id of the document each came from; follow with sabi_read_document when you need the whole thing.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/sabi"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/sabi)