Spark
Sign In
MCP Connector

Access AWS S3 Data for LLM Context

An MCP server that exposes AWS S3 buckets and PDF documents as tools and Resources for LLM context.

Connect
Works with aws s3
AWS Samples
Own this? Claim it
90
Spark score
out of 100
Updated 8 months ago
Fresher alternatives ↓
Version 1.0.0
Models
universal
Add to Favorites

Why it matters

Integrate your AWS S3 data into LLM applications. This asset provides access to S3 objects, enabling them to be loaded as resources for AI context.

Outcomes

What it gets done

01

List buckets and objects in AWS S3.

02

Retrieve specific objects from S3 buckets.

03

Format S3 data for use in LLM context.

04

Support for PDF documents within S3.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/vb-aws-s3 | bash

Capabilities

Tools your agent gets

ListBuckets

Returns a list of all buckets owned by the authenticated request sender

ListObjectsV2

Returns some or all (up to 1000) objects in a bucket with each request

GetObject

Retrieves an object from Amazon S3

Overview

AWS S3 MCP Server

An MCP server for AWS S3 that gives an LLM client bucket-listing, object-listing, and object-retrieval tools, plus PDF documents exposed as Resources. Use it when a Claude Desktop or other MCP client needs to browse S3 buckets and read PDF documents directly into its context.

What it does

An MCP server that exposes AWS S3 data as Resources for LLM context loading (currently limited to PDF documents, capped at 1000 objects) and provides three tools: ListBuckets (lists all buckets owned by the authenticated caller), ListObjectsV2 (returns up to 1,000 objects in a bucket per request), and GetObject (retrieves a specific object by its full key name, supporting both virtual-hosted-style and path-style S3 requests).

When to use - and when NOT to

Use this when an LLM client needs to answer a question grounded in a specific PDF sitting in an S3 bucket, without a human first downloading and pasting the file - the Resources mechanism loads it straight into context the way a GET endpoint would. It is not a general-purpose S3 file manager: the Resources feature currently only exposes PDF documents and is capped at 1,000 objects per listing, so it is not suited to browsing very large buckets or non-PDF file types through Resources (the ListObjectsV2/GetObject tools still work on any object type, just without the automatic context-loading Resources gives PDFs).

Capabilities

  • ListBuckets - returns every bucket owned by the authenticated sender of the request.
  • ListObjectsV2 - returns some or all (up to 1,000) of the objects in a bucket per request.
  • GetObject - retrieves a specific object by its full key name, supporting both virtual-hosted-style and path-style S3 requests.
  • Resources - exposes PDF documents from S3 directly into the LLM's context, similar to a GET endpoint, currently limited to PDFs and capped at 1,000 objects.
  • Distributed under the MIT-0 License (MIT No Attribution).

How to install

Requires an AWS access key ID, secret access key, and region with appropriate S3 permissions, obtained from the AWS Management Console. For Claude Desktop, add the server to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json; Windows: %APPDATA%/Claude/claude_desktop_config.json). The published-server configuration is:

{
  "mcpServers": {
    "s3-mcp-server": {
      "command": "uvx",
      "args": [
        "s3-mcp-server"
      ]
    }
  }
}

Since MCP servers run over stdio, debugging can be difficult - the source recommends the MCP Inspector tool for the best debugging experience, launched via npx @modelcontextprotocol/inspector.

Who it's for

Developers building Claude Desktop or other MCP-client workflows who need to browse S3 buckets and pull PDF documents into an LLM's context without writing custom S3 integration code, and who are comfortable managing AWS credentials directly.

Source README

Sample S3 Model Context Protocol Server

An MCP server implementation for retrieving data such as PDF's from S3.

Features

Resources

Expose AWS S3 Data through Resources. (think of these sort of like GET endpoints; they are used to load information into the LLM's context). Currently only PDF documents supported and limited to 1000 objects.

Tools

  • ListBuckets
    • Returns a list of all buckets owned by the authenticated sender of the request
  • ListObjectsV2
    • Returns some or all (up to 1,000) of the objects in a bucket with each request
  • GetObject
    • Retrieves an object from Amazon S3. In the GetObject request, specify the full key name for the object. General purpose buckets - Both the virtual-hosted-style requests and the path-style requests are supported

Configuration

Setting up AWS Credentials

  1. Obtain AWS access key ID, secret access key, and region from the AWS Management Console.
  2. Ensure these credentials have appropriate permissions for AWS S3.

Usage with Claude Desktop

Claude Desktop

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json
On Windows: %APPDATA%/Claude/claude_desktop_config.json

Development/Unpublished Servers Configuration 
{
  "mcpServers": {
    "s3-mcp-server": {
      "command": "uv",
      "args": [
        "--directory",
        "/Users/user/generative_ai/model_context_protocol/s3-mcp-server",
        "run",
        "s3-mcp-server"
      ]
    }
  }
}
Published Servers Configuration 
{
  "mcpServers": {
    "s3-mcp-server": {
      "command": "uvx",
      "args": [
        "s3-mcp-server"
      ]
    }
  }
}

Development

Building and Publishing

To prepare the package for distribution:

  1. Sync dependencies and update lockfile:
uv sync
  1. Build package distributions:
uv build

This will create source and wheel distributions in the dist/ directory.

  1. Publish to PyPI:
uv publish

Note: You'll need to set PyPI credentials via environment variables or command flags:

  • Token: --token or UV_PUBLISH_TOKEN
  • Or username/password: --username/UV_PUBLISH_USERNAME and --password/UV_PUBLISH_PASSWORD

Debugging

Since MCP servers run over stdio, debugging can be challenging. For the best debugging
experience, we strongly recommend using the MCP Inspector.

You can launch the MCP Inspector via npm with this command:

npx @modelcontextprotocol/inspector uv --directory /Users/user/generative_ai/model_context_protocol/s3-mcp-server run s3-mcp-server

Upon launching, the Inspector will display a URL that you can access in your browser to begin debugging.

Security

See CONTRIBUTING for more information.

FAQ

Common questions

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.