Skip to main content

MCP Resources & File Storage

MCP Resources let you give your AI access to files and data for context. This page explains what resources are, how they differ from tools, and how to use the built-in file storage.

What Are MCP Resources?​

Resources are files or data that you manually share with your AI to provide context.

Think of resources like email attachments. When you add a resource, you're giving your AI reference material to work with — like handing someone a document before asking questions about it.

Resources vs Tools: The Key Difference​

ResourcesTools
What they doProvide data to readPerform actions
Who uses themYou attach manuallyAI calls automatically
ExampleAttach a database schemaQuery the database
UIClick "+" to addAI decides when to call

Resources = Manual attachments (context)
Tools = Automatic functions (actions)

How They Work Together​

Resources and tools complement each other:

Scenario: Database Query

  1. You: Attach resource://database-schema
  2. AI: Reads it, learns table structure
  3. You: "Get all active users"
  4. AI: Automatically calls the database tool with correct table names

The resource gave knowledge. The tool performed the action. Together they enabled smart automation.

MCPBundles Vault (file storage)​

Vault is MCPBundles’ workspace file storage, exposed as MCP tools on the Vault MCP server (vault_upload, vault_list_files, vault_download, vault_delete, vault_status, and related tools). It is separate from generic MCP Resources attachments in Claude or Cursor — Vault is where your AI saves, lists, and retrieves files through MCPBundles tools after you connect the Vault server.

Get Vault working: On MCP Servers, open Directory (or search), find Vault, Connect it, and complete any connection steps on the server page. File tools do not run until the Vault MCP server is connected in your workspace.

Storage limits​

  • Max file size: 100MB per upload (enforced on Vault upload)
  • Workspace totals: Ask your AI to run Vault status tools, or use usage stats from vault_status — there is no enforced workspace-wide file count or total storage cap in the product today

What your AI can do with Vault files​

Save Files​

Your AI can save files you share or create during conversations.

Examples:

  • "Save this code snippet to my snippets folder"
  • "Take this screenshot and store it"
  • "Download this research paper and save it for me"

The AI will automatically organize files and can create subdirectories to keep things tidy.

List & Find Files​

Your AI can browse your stored files and find what you need.

Examples:

  • "What files do I have saved?"
  • "Find all my reports from this month"
  • "Show me everything in my images folder"
  • "How much storage am I using?"

The AI will show you file names, sizes, when they were last modified, and your storage usage.

Download and large files​

vault_download returns file bytes inline for smaller files. Above the inline size threshold, the tool returns a presigned S3 download URL (spillover) so your AI or client can fetch the file without loading megabytes into chat context. This is not a separate “share link” tool with configurable expiry — it is download delivery for large Vault objects.

Examples:

  • "Download my report from Vault"
  • "Get the PDF I saved last week"

Organize files​

Your AI can rename, copy, and delete files to keep your storage organized.

Examples:

  • "Rename that file to include today's date"
  • "Make a backup copy of my report"
  • "Delete all the test files"
  • "Move this to my archive folder"

Common Use Cases​

Screenshot storage​

You: "Save this screenshot to Vault"

What happens:

  • AI saves the image via vault_upload
  • You can list or download it later; large downloads may return a presigned URL

Perfect for: Debugging captures, UI mockups, visual notes you want the agent to reuse across sessions.

Research Paper Library​

You: "Download this paper and save it to my research folder"
You share: https://arxiv.org/pdf/2304.09121.pdf

What happens:

  • AI downloads the PDF from the URL
  • Saves it with an organized filename
  • You can later ask "Show me papers about LLM agents" and it will find it

Perfect for: Building a personal research library, saving articles to read later, organizing learning materials.

Code Snippet Collection​

You: "Save this function for authentication"

What happens:

  • AI saves the code snippet with a descriptive name
  • Automatically organizes it into a code folder
  • Later you can ask "Show me my auth code" and get it back

Perfect for: Reusable code patterns, solutions to common problems, team code standards.

Batch File Management​

You: "Find all my test files and rename them to include today's date"

What happens:

  • AI searches your files for anything containing "test"
  • Renames each one to add the current date
  • Confirms how many files were updated

Perfect for: Organizing downloads, archiving old files, cleaning up storage.

Using MCP Resources in your AI client​

Claude Desktop​

  1. Click the "+" button below where you type
  2. Select "Add from [your MCP server name]"
  3. Choose the resource you want to attach
  4. It appears as an attachment in your conversation
  5. Ask questions about it

VS Code / Cursor​

  1. Click "Add Context" in the chat panel
  2. Select "MCP Resources"
  3. Choose your resource
  4. It shows up in your conversation context
  5. Reference it in your questions

Other MCP Clients​

Most MCP clients have a similar pattern:

  • Look for "Add Context" or "Attach" buttons
  • Resources will appear in a list to select from
  • Once attached, your AI can read and reference them

Tips for using Vault​

Keep Files Organized​

Your AI can create folders for you - just mention where you want things saved:

  • "Save this to my images folder" → AI creates images/
  • "Put this in my 2025 reports" → AI creates 2025-reports/
  • "Add this to code snippets" → AI creates code-snippets/

Use descriptive names so you can find files later by asking natural questions.

Monitor Your Storage​

Ask your AI "How much storage am I using?" and it can report Vault usage (file counts and bytes used via Vault status tools):

Keep downloads private​

Presigned download URLs from vault_download are time-limited and meant for fetching a file once. Do not post them in public channels. Vault files stay workspace-scoped like other MCPBundles data.

Clean up regularly​

Ask your AI to help keep storage tidy:

  • "Delete files I haven't used in 6 months"
  • "Remove all test files"
  • "Show me my largest files"

Troubleshooting​

File Too Large​

If a file is over 100MB, you can:

  • Compress it first (zip, reduce image quality, etc.)
  • Split it into smaller parts
  • Store it elsewhere and save just the link

Freeing Space​

To reduce Vault usage:

  1. Ask "What are my largest files?"
  2. Delete old or unnecessary files
  3. Archive large files to external storage (Google Drive, Dropbox)
  4. Keep only actively needed files in Vault

URL Download Fails​

If your AI can't download from a URL:

  • Make sure the URL works in your browser
  • Check if it requires login (AI can't handle auth)
  • Try a direct link instead of a redirect URL
  • Some sites block automated downloads - download manually instead

Next Steps​