rustfs/crates/mcp/README.md

[![RustFS](https://rustfs.com/images/rustfs-github.png)](https://rustfs.com)

# RustFS MCP Server - Model Context Protocol

<p align="center">
  <strong>High-performance MCP server providing S3-compatible object storage operations for AI/LLM integration</strong>
</p>

<p align="center">
  <a href="https://github.com/rustfs/rustfs/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/rustfs/rustfs/actions/workflows/ci.yml/badge.svg" /></a>
  <a href="https://docs.rustfs.com/">📖 Documentation</a>
  <a href="https://github.com/rustfs/rustfs/issues">🐛 Bug Reports</a>
  <a href="https://github.com/rustfs/rustfs/discussions">💬 Discussions</a>
</p>

---

## 📖 Overview

**RustFS MCP Server** is a high-performance [Model Context Protocol (MCP)](https://spec.modelcontextprotocol.org) server that provides AI/LLM tools with seamless access to S3-compatible object storage operations. Built with Rust for maximum performance and safety, it enables AI assistants like Claude Desktop to interact with cloud storage through a standardized protocol.

### What is MCP?

The Model Context Protocol is an open standard that enables secure, controlled connections between AI applications and external systems. This server acts as a bridge between AI tools and S3-compatible storage services, providing structured access to file operations while maintaining security and observability.

## ✨ Features

### Supported S3 Operations

- **List Buckets**: List all accessible S3 buckets
- **List Objects**: Browse bucket contents with optional prefix filtering
- **Upload Files**: Upload local files with automatic MIME type detection and cache control
- **Get Objects**: Retrieve objects from S3 storage with read or download modes

## 🔧 Installation

### Prerequisites

- Rust 1.70+ (for building from source)
- AWS credentials configured (via environment variables, AWS CLI, or IAM roles)
- Access to S3-compatible storage service

### Build from Source

```bash
# Clone the repository
git clone https://github.com/rustfs/rustfs.git
cd rustfs

# Build the MCP server
cargo build --release -p rustfs-mcp

# The binary will be available at
./target/release/rustfs-mcp
```

## ⚙️ Configuration

### Environment Variables

```bash
# AWS Credentials (required)
export AWS_ACCESS_KEY_ID=your_access_key
export AWS_SECRET_ACCESS_KEY=your_secret_key
export AWS_REGION=us-east-1  # Optional, defaults to us-east-1

# Optional: Custom S3 endpoint (for MinIO, etc.)
export AWS_ENDPOINT_URL=http://localhost:9000

# Logging level (optional)
export RUST_LOG=info
```

### Command Line Options

```bash
rustfs-mcp --help
```

The server supports various command-line options for customizing behavior:

- `--access-key-id`: AWS Access Key ID for S3 authentication
- `--secret-access-key`: AWS Secret Access Key for S3 authentication
- `--region`: AWS region to use for S3 operations (default: us-east-1)
- `--endpoint-url`: Custom S3 endpoint URL (for MinIO, LocalStack, etc.)
- `--log-level`: Log level configuration (default: rustfs_mcp_server=info)

## 🚀 Usage

### Starting the Server

```bash
# Start the MCP server
rustfs-mcp

# Or with custom options
rustfs-mcp --log-level debug --region us-west-2
```

### Integration with chat client

#### Option 1: Using Command Line Arguments

```json
{
  "mcpServers": {
    "rustfs-mcp": {
      "command": "/path/to/rustfs-mcp",
      "args": [
        "--access-key-id", "your_access_key",
        "--secret-access-key", "your_secret_key",
        "--region", "us-west-2",
        "--log-level", "info"
      ]
    }
  }
}
```

#### Option 2: Using Environment Variables

```json
{
  "mcpServers": {
    "rustfs-mcp": {
      "command": "/path/to/rustfs-mcp",
      "env": {
        "AWS_ACCESS_KEY_ID": "your_access_key",
        "AWS_SECRET_ACCESS_KEY": "your_secret_key",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}
```

### Using MCP with Docker

#### Docker image build

Using MCP with docker will simply the usage of rustfs mcp. Building the docker image with below command:

```
docker build -f Dockerfile -t rustfs/rustfs-mcp ../../
```

Alternatively, if you want to build the image from the rustfs codebase root directory,run the command:

```
docker build -f crates/mcp/Dockerfile -t rustfs/rustfs-mcp .
```

#### IDE Configuration

Adding the following content in IDE MCP settings:

```
{
  "mcpServers": {
    "rustfs-mcp": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "AWS_ACCESS_KEY_ID",
        "-e",
        "AWS_SECRET_ACCESS_KEY",
        "-e",
        "AWS_REGION",
        "-e",
        "AWS_ENDPOINT_URL",
        "rustfs/rustfs-mcp"
      ],
      "env": {
        "AWS_ACCESS_KEY_ID": "rustfs_access_key",
        "AWS_SECRET_ACCESS_KEY": "rustfs_secret_key",
        "AWS_REGION": "cn-east-1",
        "AWS_ENDPOINT_URL": "rustfs_instance_url"
      }
    }
  }
}
```

If success, MCP configure page will show the [available tools](#️-available-tools).

## 🛠️ Available Tools

The MCP server exposes the following tools that AI assistants can use:

### `list_buckets`

List all S3 buckets accessible with the configured credentials.

**Parameters:** None

### `list_objects`

List objects in an S3 bucket with optional prefix filtering.

**Parameters:**

- `bucket_name` (string): Name of the S3 bucket
- `prefix` (string, optional): Prefix to filter objects

### `upload_file`

Upload a local file to S3 with automatic MIME type detection.

**Parameters:**

- `local_file_path` (string): Path to the local file
- `bucket_name` (string): Target S3 bucket
- `object_key` (string): S3 object key (destination path)
- `content_type` (string, optional): Content type (auto-detected if not provided)
- `storage_class` (string, optional): S3 storage class
- `cache_control` (string, optional): Cache control header

### `get_object`

Retrieve an object from S3 with two operation modes: read content directly or download to a file.

**Parameters:**

- `bucket_name` (string): Source S3 bucket
- `object_key` (string): S3 object key
- `version_id` (string, optional): Version ID for versioned objects
- `mode` (string, optional): Operation mode - "read" (default) returns content directly, "download" saves to local file
- `local_path` (string, optional): Local file path (required when mode is "download")
- `max_content_size` (number, optional): Maximum content size in bytes for read mode (default: 1MB)

### `create_bucket`

Create a new S3 bucket with the specified name.

**Parameters:**

- `bucket_name` (string): Source S3 bucket.

### `delete_bucket`

Delete the specified S3 bucket. If the bucket is not empty, the deletion will fail. You should delete all objects and objects inside them before calling this method.**WARNING: This operation will permanently delete the bucket and all objects within it!**

- `bucket_name` (string): Source S3 bucket.

## Architecture

The MCP server is built with a modular architecture:

```
rustfs-mcp/
├── src/
│   ├── main.rs          # Entry point, CLI parsing, and server initialization
│   ├── server.rs        # MCP server implementation and tool handlers
│   ├── s3_client.rs     # S3 client wrapper with async operations
│   ├── config.rs        # Configuration management and CLI options
│   └── lib.rs           # Library exports and public API
└── Cargo.toml           # Dependencies, metadata, and binary configuration
```