s3-mcp

s3-mcp

Enables managing and interacting with any S3-compatible storage endpoint through MCP tools for listing, reading, writing, copying, and deleting buckets and objects.

Category
访问服务器

README

s3-mcp

A generic S3-protocol MCP server (stdio) that works against any S3-compatible endpoint — RustFS, MinIO, Cloudflare R2, Backblaze B2, AWS S3 — not just AWS.

Single Python package, one boto3 client created at startup, configuration via environment variables only. Built on the official mcp SDK (high-level server API) and boto3.

Configuration (environment variables)

Variable Required Default Description
S3_ENDPOINT_URL no (unset) S3-compatible endpoint; omit to target real AWS
AWS_REGION no us-east-1 Region for signing and bucket creation
AWS_ACCESS_KEY_ID yes — Access key
AWS_SECRET_ACCESS_KEY yes — Secret key
S3_PATH_STYLE no true Path-style addressing (https://host/bucket/key) — required by most non-AWS stores
S3_TLS_INSECURE no false true disables TLS certificate verification (self-signed endpoints)

Booleans accept true/false/1/0/yes/no (case-insensitive). Missing credentials or invalid booleans abort startup with a clear error.

Tools

Tool Read-only hint
list_buckets() yes
list_objects(bucket, prefix="", max_keys=1000) yes
get_object(bucket, key) → text if UTF-8 else {base64, content_type} yes
stat_object(bucket, key) yes
presign_get(bucket, key, expires_s=3600) yes
put_object(bucket, key, body, base64=false) no
delete_object(bucket, key) no
copy_object(src_bucket, src_key, dst_bucket, dst_key) no
create_bucket(bucket) no
delete_bucket(bucket) no
presign_put(bucket, key, expires_s=3600) no

Tool errors are raised as short RuntimeError messages and surfaced by MCP as error results. Presigned URLs are capped at 7 days (604800 s), per SigV4.

Example configurations

RustFS with a self-signed certificate

{
  "mcpServers": {
    "s3-mcp": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "S3_ENDPOINT_URL=https://rustfs.local:9000",
        "-e", "S3_TLS_INSECURE=true",
        "-e", "AWS_REGION=us-east-1",
        "-e", "AWS_ACCESS_KEY_ID",
        "-e", "AWS_SECRET_ACCESS_KEY",
        "ghcr.io/jakeperalta7/s3-mcp:latest"
      ]
    }
  }
}

-e VAR without a value passes the variable through from your shell, so credentials never appear in the config file.

MinIO

{
  "mcpServers": {
    "s3-mcp": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "S3_ENDPOINT_URL=http://localhost:9000",
        "-e", "AWS_REGION=us-east-1",
        "-e", "AWS_ACCESS_KEY_ID",
        "-e", "AWS_SECRET_ACCESS_KEY",
        "ghcr.io/jakeperalta7/s3-mcp:latest"
      ]
    }
  }
}

AWS S3

Omit S3_ENDPOINT_URL; standard AWS credential/region resolution applies:

{
  "mcpServers": {
    "s3-mcp": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "AWS_REGION",
        "-e", "AWS_ACCESS_KEY_ID",
        "-e", "AWS_SECRET_ACCESS_KEY",
        "ghcr.io/jakeperalta7/s3-mcp:latest"
      ]
    }
  }
}

Without Docker (uv)

uvx --from git+https://github.com/JakePeralta7/s3-mcp s3-mcp
{
  "mcpServers": {
    "s3-mcp": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/JakePeralta7/s3-mcp", "s3-mcp"],
      "env": {
        "S3_ENDPOINT_URL": "http://localhost:9000",
        "AWS_ACCESS_KEY_ID": "...",
        "AWS_SECRET_ACCESS_KEY": "..."
      }
    }
  }
}

Development

uv sync --group dev   # install deps into .venv
uv run pytest -q      # unit tests (boto3 mocked, no network)
docker build -t s3-mcp:dev .

Entrypoint: python -m s3_mcp (stdio transport only).

Releasing

Every push to main runs tests, then publishes ghcr.io/<owner>/<repo>:<version> + :latest, where <version> is read from version in pyproject.toml. To release, bump the version and merge to main. Git tags are not used.

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选