manifest-mcp
Enables AI assistants to read and update a user's personal manifest (portable context) through a secure, token-based connector.
README
manifest-mcp
The server behind the link option at readthemanifest.net — one personal connector URL that lets Claude and ChatGPT read your manifest (your portable context) and save new facts to it.
This repo exists so you don't have to take our word for what the server does. Read it. It's three small files.
How it treats your manifest
Your connector URL looks like https://readthemanifest.net/mcp/<token>. The token is a
long random secret, and it does two jobs:
- It's the address. The server hashes the token (SHA-256) and uses the hash to label your storage slot. The token itself is never written down — not in a database, not in a log we keep.
- It's the key. Your manifest is encrypted (AES-256-GCM) with a key derived from the token (HKDF). So what sits in storage is a locked file the server cannot open on its own. It is only unlocked in the moment your AI asks for it, because the request carries the key.
No accounts, no email, no passwords. The honest consequence: lose the URL and the manifest is unrecoverable — there is nothing to reset. Export a copy anytime.
What "open source" does and doesn't prove
Reading this code tells you what the published server does. It cannot prove what a remote server is running — no open-source project's deployment can. If that distinction matters to you, Manifest's GitHub options keep your context entirely in your own hands, and this repo lets you run your own copy of this server (below) so the deployment is yours too.
The files
netlify/functions/mcp.mjs— the MCP endpoint (POST /mcp/:token). Stateless JSON-RPC over HTTP. Two tools:get_manifest(read) andadd_to_manifest(append one dated fact).netlify/functions/manifest-vault.mjs— vault management (POST /api/manifest-vault): create, replace, export, delete.netlify/functions/_lib/vault.mjs— the storage and crypto described above. Storage is Netlify Blobs.
Limits: manifest ≤ 150KB, note ≤ 4KB.
Run your own
Deploy this repo to any Netlify account (free tier works; Blobs is enabled automatically),
then add your own https://<your-site>/mcp/<token> as a custom connector. To try it
locally:
npm install
npx netlify dev --offline
# create a vault
curl -s -X POST http://localhost:8888/api/manifest-vault \
-H 'Content-Type: application/json' \
-d '{"action":"create","manifest":"# Manifest\n\nHello."}'
# read it the way your AI would (use the returned token)
curl -s -X POST http://localhost:8888/mcp/<token> \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_manifest","arguments":{}}}'
(Offline local dev stores vaults as files under .netlify/local-blobs/; deployed, it's
Netlify Blobs with strong read-after-write consistency.)
License
MIT.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。