What Is MCP? Why Note Apps Are Changing in the Age of AI Tool Use
Key takeaways
- MCP is a standard way for AI clients to connect to external tools and data.
- When a note app supports MCP, it can become persistent memory that AI reads and writes during chat.
- Jade Note uses MCP with OAuth and scoped tools so AI access can be useful and controlled.
We’ve entered the era where ChatGPT and Claude can reach directly into your tools. The key that makes this possible is MCP — the Model Context Protocol.
MCP in a Nutshell
MCP is an open standard protocol for connecting AI (LLMs) to external tools and data sources. Introduced by Anthropic, it’s now supported by major AI clients including Claude Desktop and ChatGPT.
Think of MCP as a USB port for AI:
- MCP servers provide tools and data — calendars, databases, and yes, note apps
- MCP clients are the AI side that uses them — Claude Desktop, ChatGPT, and others
A service that supports MCP doesn’t need to build a separate plugin per AI vendor. Publish one server, and every MCP-compatible client can use it.
What MCP Changes
AI chat used to be a sealed box: the only way to give it your data was copy-paste. With MCP, the AI calls tools naturally, mid-conversation:
User: “From last month’s reading notes, pick the three books that stuck with me.” AI: (calls the note app’s search tool, answers from the results)
You never switch apps. The AI uses the tool behind the scenes and answers in context.
Note Apps Become AI Memory
One of MCP’s biggest consequences is a role change for note apps.
LLMs have no memory across conversations. But if there’s a note app the AI can read and write via MCP, that app becomes the AI’s persistent memory. Each conversation, the AI searches your notes to recover context and writes down what it learns. The note app shifts from “a human’s recording tool” to “a memory shared with AI.”
A Real Example: Jade Note’s MCP Server
Jade Note is an AI memory layer built around exactly this idea. It authenticates securely with OAuth 2.0 (PKCE) and exposes 22 MCP tools to AI clients:
- Reading —
search_notes(keyword + semantic),get_note,list_categories,get_category_wiki, and more - Writing —
create_note,append_to_note,patch_note,create_note_link, and more - Safety —
preview_note_updateto review changes before they land, automatic versioning, andrestore_note_versionfor rollback - Organization & permissions — per-category purpose descriptions guide the AI to the right destination, and per-category access levels (read-write, read-only, or blocked) give fine-grained control over what the AI can see and touch
The design assumes autonomous AI writes from day one, which is why change previews and version restore are built in. And because the product is built to be operated over MCP, tool responses stay compact — keeping round trips fast and token consumption low. If “letting an AI manage my memory” sounds risky, this is the answer — at both the protocol and the product level.
Takeaways
- MCP is the standard protocol connecting AI to external tools
- One server works across all compatible clients (Claude Desktop, ChatGPT, …)
- Through MCP, note apps gain a new role: the AI’s persistent memory
Try the MCP-connected note experience yourself.
Start free — includes a 60-day Pro Trial