
Geniffy
io.github.Geniffyv1.0.0更新於 Oct 5, 2026
Your memory in every AI app: each fact with the line it came from, or a plain "nothing stored".
概覽
一個託管記憶服務,讓已連線的 AI 應用搜尋、讀取、儲存、更正與遺忘你告訴它們的事實,並附上來源行與日期。
- 功能
- Geniffy 是遠端 MCP 伺服器,為 AI 應用提供共用記憶。工具包括 search 與 fetch,用來找出事實及其來源行;ask 只依據已存記憶作答;profile 給出長期成立的事實;list_memories 依類型瀏覽;remember、correct 與 forget 用來修改已存內容。事實帶有日期,較新的事實會取代較舊的。當沒有依據支持答案時,伺服器會直接說明,而不是猜測。
- 適用情境
- 當你希望 Claude、ChatGPT、Cursor、VS Code 或 Codex 等多個支援 MCP 的應用共用同一份記憶,且答案可追溯到來源行時,值得加入。若只需在單一應用內記憶,或不願把筆記存放在第三方,則幫助不大。
- 執行需求
- 遠端 Streamable HTTP 端點,不需本機安裝。內建 OAuth 2.1 登入,多數應用會開啟瀏覽器登入並由你核准授權。指令碼、CI 及無法開啟瀏覽器的代理可改為在 Authorization 標頭中傳送 Geniffy API 金鑰。需要能連線至該託管服務的網路。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Geniffy,將其新增到工作區。
- 為需要使用其工具的對話啟用該服務。
Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
其他 MCP 客戶端
把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"geniffy-mcp": {
"type": "http",
"url": "https://api.geniffy.com/mcp"
}
}
}README
Geniffy MCP server
Your memory inside Claude, ChatGPT, Cursor, VS Code, Codex and any app that speaks MCP. Tell one of them something, and every app you connect can find it, with the line it came from and the date it was said.
[Docs] [Install in VS Code] [Add to Cursor] [License: MIT]
A hosted, remote MCP server (Streamable HTTP) with sign-in built in: there is nothing to install and no key to copy. Add the URL to your app, sign in to Geniffy when it asks, and choose what the app may do.
Listed in the official MCP Registry
as io.github.Geniffy/geniffy-mcp.
Connect your app
Cursor, in ~/.cursor/mcp.json:
VS Code, in .vscode/mcp.json:
Tools
Every tool has an input schema, an output schema and annotations that say whether it changes your memory.
search and fetch return the shape ChatGPT's connectors read. Inputs and results:
docs.geniffy.com/mcp/tools.
How it answers
- With its source. Every fact comes back with the line it was read from, who said it and when.
- Or not at all. When nothing in your memory supports an answer, Geniffy says so, and the app is told to say it doesn't know rather than guess.
- With today's truth. Facts carry dates. When something changes, the newer fact replaces the older one.
What you control
- You approve each app. Geniffy shows who is asking and where it will send you back. Nothing is shared until you press Allow.
- Read, or read and change. Turn off saving when you connect, and the app never sees the tools that change your memory.
- Disconnect at any time in the Geniffy app under Agents and MCP. Every call an app makes is listed under Requests, with its name.
- Your memory stays in India, where Geniffy keeps it.
Sign-in, for client developers
OAuth 2.1, following the MCP authorization specification: a 401 with WWW-Authenticate names the
protected-resource metadata (RFC 9728), which points to the authorization server metadata (RFC 8414).
Dynamic client registration (RFC 7591), PKCE with S256 only, scopes memory:read and memory:write,
hourly access tokens, rotating refresh tokens, revocation (RFC 7009) and the iss response parameter
(RFC 9207). Details: docs.geniffy.com/mcp/sign-in.
Without a browser
Scripts, CI and agents that cannot open a browser can send a Geniffy API key as a header. A key reaches
your app's spaces too: every tool then takes an optional space.
Build your own app on Geniffy
The MCP server gives the AI apps you use your memory. To give your own app's users a memory, use the API:
- Python:
pip install geniffy - TypeScript and JavaScript:
npm install geniffy - Runnable examples: Geniffy/examples
- Docs: docs.geniffy.com
Security
Report a vulnerability to [email protected]. Please don't open a public issue for it.
來源:README.md,提交 9e9588e
工具
0版本歷史
1- v1.0.0最新Oct 5, 2026


