How to connect Claude to your notes

Claude is good at the slow kind of thinking: working through a decision, drafting something long, or making sense of a pile of notes. All of that goes better when Claude can see what you already know, and when the result does not vanish when you close the chat.

You can connect Claude to your notes in two ways. In the Claude app, Claude Desktop, and Cowork, you add a connector from Claude's directory. In Claude Code, you add the same connection with one command. After that, Claude can search your notes, answer from them, write new ones, and edit old ones.

We make Capacities, so this guide uses it. For other notes apps that work with Claude, see our list of note-taking apps for AI.

The short answer

You pick one space when you sign in. You need Capacities Pro. The setup guide has screenshots.

What Claude can do with your notes

Answer from what you have written

Claude searches your space, reads the notes that match, and answers from them. It links to the notes it used.

• "What did I decide about the new pricing, and why?"
• "Read my notes on Botticelli and compare the themes across his paintings in a table."
• "Which people have I met through the design conference? What do I know about each?"

Make sense of many notes

This is where Claude shines. Point it at a set of notes and ask what it sees.

Ask it to narrow first, then read. "List my Interview notes from September" works better than "read all my notes". Claude picks up on this quickly when you say it once.

Write and edit

• "Save the outline we just made as a page called 'Essay on slow productivity', linked to my Writing page."
• "Add a Person object for today's podcast guest, with their name and role, and put my notes in the body."
• "Shorten the second paragraph of my draft and keep my voice."
• "Mark the Book 'The Overstory' as finished and add today's date."

Every edit is saved in the note's version history.

Cowork and longer tasks

In Cowork, Claude can work through a longer task step by step while you do something else. With your notes connected, that task can start from your context and end in your notes.

• "Read my meeting notes from this week, and write a one-page summary for Friday into today's daily note."
• "Go through my reading notes tagged #habits and draft a blog post outline from them."

Some people set up a scheduled task in Claude that writes a short briefing into their daily note every morning. The connection works the same way for those.

Claude Code

Claude Code is Claude in your terminal. Developers use it to write and change code, and more and more people use it for writing and research too.

Add the connection once:

claude mcp add --transport http capacities https://api.capacities.io/mcp

Then type /mcp in Claude Code and sign in. Use /context to see how much room each connected server takes in the context window.

What it is good for:

• Decisions next to the code. "Read my Capacities note on the billing redesign before you start." The why stays in your notes, and the code stays in the repository.
• Writing back what happened. "Write a short note in Capacities about what we changed today and why, linked to the Billing project."
• Turning research into a draft. "Pull my notes tagged #onboarding and draft the docs page from them."

If you want Claude Code to build a whole knowledge base for you and your agents, read our guide to an AI second brain. It covers Cursor, Codex, and when a folder of Markdown files is the better choice.

Claude memory and Projects are not a replacement

Claude has two features that sound similar.

Memory is a summary Claude writes about you from your conversations, updated every 24 hours. Anthropic says it focuses on things like your role, your projects, and your preferences. It helps Claude know you. It is not a place where the insight itself lives as a page you can open, change, and link.

Projects hold files and instructions for one effort, and each project has its own memory. That keeps projects tidy, and it also means what you learned in one project does not show up in the next. On the free plan you can have up to five projects.

Your notes sit next to both. They hold what you know across every project, they are pages you control, and every assistant you connect reads the same ones.

Privacy and control

• Claude reaches only the space you picked, and reads only what it looks up for your request.
• You can see and remove each connection under Settings > AI Assistant in Capacities. Remove it in Claude as well.
• Claude asks before it uses a tool, unless you allow it to always use that tool. You can switch tools to ask every time in the connector settings.
• Anthropic's data settings apply to what Claude reads. Check the privacy settings in your Claude account.
• Capacities stores your notes in the EU. See our AI privacy page.

What it costs

Connecting Claude needs Capacities Pro. Your Claude plan pays for the AI. Chats in Claude do not use your Capacities AI budget.

Who should not use this

• If you want Claude Code to work only on files on your disk. Claude Code is excellent at reading and rewriting a folder of Markdown files. Obsidian or Logseq keep your notes in that form. Capacities is not file-based, although you can export to Markdown at any time.
• If your team needs one shared knowledge base. Capacities is for one person.
• If you want Claude to design your structure. It cannot create object types or properties yet.

How we wrote this guide

• We make Capacities. Claims about it come from our setup guide and MCP documentation.
• Facts about Claude come from Anthropic's help pages, read in October 2026: connectors, custom connectors, memory, and Projects.
• If something here is wrong, email [email protected].