What You Can Upload to Claude: A Practical Guide to Making Claude Remember

Claude has seven distinct ways to store and access information. Some persist forever, some last one conversation, and some connect to live external data. This guide covers each one: what it is, how to use it, and when it matters.

Note: This guide reflects the full Claude ecosystem, including Claude.ai, the Claude desktop app, Claude Code, and Claude for Work. Some layers (like CLAUDE.md and workspace folders) apply primarily to Claude Code and desktop workflows.

At a Glance

Layer	Persists?	Who Controls It	Best For
Memory Files	Across all sessions	Claude + You	Facts about you, your clients, preferences
CLAUDE.md	Across all sessions	You (or Claude)	Hard rules, formatting standards, identity
Skills	Across all sessions	You + Plugins	Repeatable workflows, content frameworks
Plugins	Across all sessions	Marketplace + You	Bundled skill sets, MCP connectors
Workspace Folder	On your computer	You	Project files, deliverables, reference docs
File Uploads	One conversation	You	Quick reference, docs to edit, images
MCP Connectors	Across all sessions	You (connect once)	Live data from Gmail, Slack, Calendar, etc.

1. Memory Files

Memory is Claude's long-term knowledge base. It stores facts, preferences, feedback, and reference information in individual markdown files, indexed by a master file called MEMORY.md.

What gets stored here

User memories: who you are, how you work, your role, your preferences.

Feedback memories: corrections you've given Claude ("never put timelines in docs").

Project memories: ongoing initiatives, client details, decisions made.

Reference memories: where to find things in external systems (Slack channels, Granola folders, spreadsheet links).

How to add to it

Say "remember that..." and Claude saves it immediately.

Claude also saves things proactively when it learns something important during a conversation.

You can ask Claude to read, update, or delete any memory file.

How to review it

Ask "what's in my memory?" or "read my memory index" and Claude will show you the full list. You can also ask for specific files: "read the CORE40 team directory memory."

Tip: Memory is the most powerful persistence layer you have. If you notice Claude forgetting something between sessions, tell it to save that thing to memory. One sentence from you creates a permanent record.

2. CLAUDE.md (Global Instructions)

CLAUDE.md is a file that loads at the start of every single conversation. It contains hard rules that override default behavior. Think of it as Claude's operating manual for working with you.

What belongs here

Formatting rules (your .docx standards, font preferences, timestamp format).

Identity rules ("My brand is Human First Media, never 3 Outcomes").

Output rules ("All documents opened in Google Docs").

Behavior rules ("Always ask before including pricing").

How to edit it

Say "update my global instructions to include..." or "add this rule to my CLAUDE.md." Claude can edit it directly. You can also ask to see the current contents.

CLAUDE.md vs. Memory

CLAUDE.md is for rules. Memory is for facts. If something is a non-negotiable instruction ("use Arial, always"), it goes in CLAUDE.md. If something is context Claude should know ("Pete is a civil engineer"), it goes in memory.

3. Skills

Skills are reusable instruction sets that tell Claude how to do a specific type of work. You have over 60 of them. When you say "write a LinkedIn article," Claude loads the medland-linkedin-article skill, which contains your exact editorial format, voice rules, and structure.

Where skills come from

Pre-built skills that come with Cowork (docx, pptx, pdf, xlsx).

Skills bundled inside plugins you install (the marketing plugin, the legal plugin, the brand-voice plugin).

Custom skills you or Claude create specifically for your workflows (dean-mode, no-ai-voice, content-carol).

How to create a new skill

Say "create a skill for [workflow]." Claude uses the skill-creator skill to build a new one. You describe the workflow, and Claude writes the instruction file, sets trigger phrases, and registers it in your library.

How to see what you have

Say "what skills do I have" or "list my skills" and Claude renders the full catalog.

Tip: Skills are the reason your output quality stays consistent across sessions. Without them, Claude would start from scratch every time. If you find yourself repeating the same instructions, that's a sign it should be a skill.

4. Plugins

Plugins are installable bundles that package skills, MCP connectors, and tools together. Your marketing plugin, for example, bundles SEO audit, campaign planning, email sequences, and competitive briefs into one install.

How to install

Say "search plugins for [topic]" and Claude searches the marketplace.

Claude recommends relevant matches and you approve the install.

Some plugins require authentication (connecting your Slack, Gmail, etc.).

What you have installed

Ask "list my plugins" to see everything currently installed. Right now you have plugins for brand voice, marketing, legal, and Cowork plugin management, among others.

5. Workspace Folder

When you start a Cowork session, you can select a folder from your computer. Claude can read from and write to that folder. Your current workspace is the MABEL project folder.

What to put here

Reference documents you want Claude to access (brand guides, client briefs, contracts).

Files you want Claude to edit or update.

Nothing sensitive you wouldn't want an AI assistant to see.

How it works

Claude sees everything in the folder you selected. Any files Claude creates for you (documents, spreadsheets, presentations) get saved here too, so you can open them directly from Finder or Google Drive.

Tip: If you want Claude to work with files on your computer that aren't in the current workspace, you can ask Claude to request a different folder. You can also drag and drop files into the conversation (see File Uploads below).

6. File Uploads (Per-Conversation)

You can drag and drop files directly into any conversation. Claude can read PDFs, images, Word docs, spreadsheets, text files, and more.

What to know

Uploaded files only last for that one conversation. They don't carry over to the next session.

If you want Claude to remember something from an uploaded file permanently, ask Claude to save the key details to memory.

Claude can see images natively (screenshots, photos, diagrams) and read text from them.

For large PDFs, Claude reads them in page ranges.

Common uses

"Here's a contract, review it" (drag in the PDF).

"Here's a screenshot of what I'm seeing" (drag in the image).

"Edit this document" (drag in the .docx).

7. MCP Connectors (Live External Data)

MCP connectors let Claude pull live data from your tools without you having to copy-paste anything. You authenticate once, and Claude can read from that service in every future session.

What you have connected

Connector	What Claude Can Do With It
Gmail	Search threads, read emails, draft replies
Google Calendar	List events, create events, find free time
Google Drive	Search files, read documents, create files
Slack	Read channels, search messages, send messages
Granola	Search meetings, pull transcripts
Zoom	List recordings, search meetings
iMessages	Read and send iMessages, search contacts
Apple Notes	Read, create, and update notes
Canva	Generate designs, search designs, export
Vercel	Deploy sites, check deployments, read logs

How to add more

Say "I need to connect [tool name]" and Claude will search the MCP registry for available connectors. If one exists, Claude walks you through the authentication flow.

Quick Reference: What to Say

Here are the exact phrases that trigger each action.

You Want To...	Say This
Save a fact permanently	"Remember that [fact]"
See what Claude remembers	"What's in my memory?"
Add a global rule	"Add to my global instructions: [rule]"
See your global rules	"Show me my CLAUDE.md"
Create a reusable workflow	"Create a skill for [workflow]"
See available skills	"List my skills"
Find new plugins	"Search plugins for [topic]"
Connect a new tool	"Connect my [Slack/Gmail/etc.]"
Work with files on your Mac	"I need to work with files in [folder]"
Delete a memory	"Forget the memory about [topic]"
Schedule a recurring task	"Schedule [task] to run every [frequency]"

The Mental Model

Think of it in layers, from most permanent to least permanent:

CLAUDE.md (hard rules, loads every session, you control it).

Memory files (facts and context, loads every session, Claude + you control it).

Skills and plugins (repeatable workflows, always available, triggered by what you say).

MCP connectors (live data from external tools, authenticate once).

Workspace folder (files on your computer, available while selected).

File uploads (one conversation only, drag and drop).

The more you invest in the top layers, the less you have to repeat yourself. That's the whole game.