Skip to main content
Glama
S3 MCP Server

S3 MCP Server

by Geun-Oh

S3 MCP Server

A Model Context Protocol (MCP) server for accessing Amazon S3 buckets. This server provides seamless integration with S3 storage through MCP, allowing efficient handling of large files including PDFs through streaming capabilities.

Features

  • S3 bucket object listing with prefix filtering

  • Efficient large file handling through streaming

  • Secure AWS credentials management

  • TypeScript support

  • CLI interface with customizable options

Related MCP server: MCP Server for Apache OpenDAL™

Installation

npx -y @geunoh/s3-mcp-server

Usage

Command Line Options

npx -y @geunoh/s3-mcp-server [options]

Options:

  • --port, -p: Server port (default: 3000)

  • --region, -r: AWS region (default: ap-northeast-2)

  • --bucket, -b: S3 bucket name (default: my-dancing-bucket)

  • --content-type, -t: Input file content type for uploads/downloads (default: application/octet-stream)

Environment Variables

Required:

export AWS_ACCESS_KEY_ID="your_access_key"
export AWS_SECRET_ACCESS_KEY="your_secret_key"

Optional:

export AWS_REGION="ap-northeast-2"
export S3_BUCKET_NAME="my-bucket-name"
export CONTENT_TYPE="application/octet-stream" # Optional: default MIME type for uploads/downloads

MCP Integration

Add to your mcp.json:

{
  "mcpServers": {
    "s3-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "@geunoh/s3-mcp-server",
        "--region",
        "us-east-1",
        "--bucket",
        "my-test-bucket",
        "--content-type",
        "text/plain"
      ],
      "env": {
        "AWS_ACCESS_KEY_ID": "YOUR_AWS_ACCESS_KEY_ID",
        "AWS_SECRET_ACCESS_KEY": "YOUR_AWS_SECRET_ACCESS_KEY"
      }
    }
  }
}

Or like this way:

{
  "mcpServers": {
    "s3-mcp-server": {
      "command": "npx",
      "args": ["-y", "@geunoh/s3-mcp-server"],
      "env": {
        "AWS_ACCESS_KEY_ID": "YOUR_AWS_ACCESS_KEY_ID",
        "AWS_SECRET_ACCESS_KEY": "YOUR_AWS_SECRET_ACCESS_KEY",

        // optional
        "AWS_REGION": "us-east-1",
        "S3_BUCKET_NAME": "my-test-bucket"
      }
    }
  }
}

Available MCP Functions

listObjects

Lists objects in the S3 bucket.

Parameters:

  • prefix (optional): Filter objects by prefix

getObject

Retrieves an object from the S3 bucket. Optimized for large files through streaming.

Parameters:

  • key: The key of the object to retrieve

Returns:

  • stream: ReadableStream of the object content

  • contentType: MIME type of the object

  • contentLength: Size of the object in bytes

  • lastModified: Last modification timestamp

  • text: Text buffer of raw pdf ByteArray

AWS IAM Permissions

Minimum required permissions (see s3-policy.json):

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": ["s3:ListBucket", "s3:GetObject"],
      "Resource": "arn:aws:s3:::my-bucket-name"
    }
  ]
}

Development

  1. Clone the repository:

git clone https://github.com/Geun-Oh/s3-mcp-server.git
cd s3-mcp-server
  1. Install dependencies:

npm install
  1. Build the project:

npm run build
  1. Run locally:

node dist/cli.js

Project Structure

.
├── src/              # TypeScript source files
├── dist/            # Compiled JavaScript files and runtime dependencies
├── tsconfig.json    # TypeScript configuration
└── package.json     # Project configuration and dependencies

Deployment

  1. Create a new version tag:

npm version patch
  1. Push to npm registry:

npm publish --access public

The GitHub Actions workflow will automatically publish the package when a new version tag is pushed.

License

MIT

Contributing

Issues and pull requests are welcome. Please ensure that your changes maintain the existing code style and include appropriate tests.

Available Tools

3 tools
get_objectD

Inputs: key (required).

list_bucketsD

Inputs: prefix (required).

upload_fileD

Inputs: localFilePath (required), key (required), contentType.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    D1.9/5.0

    Scored across 3 tools

    Disambiguation5/5

    Each tool targets a distinct S3 action (get object, list buckets, upload file) with no overlap. Even without descriptions, the names clearly differentiate the operations.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern with snake_case (get_object, list_buckets, upload_file), making them predictable and easy to understand.

    Tool Count4/5

    With 3 tools, the server is slightly under-scoped for a typical S3 service, but it covers core operations. The count is still within a reasonable range for a focused utility.

    Completeness2/5

    The tool set is missing essential S3 operations like delete_object, list_objects, and copy_object. These omissions create significant gaps that would hinder an agent performing common storage tasks.

    Related MCP Connectors

    Related MCP Servers