Access Qiniu Cloud Storage and Services

Qiniu MCP Server is built on Qiniu Cloud products and lets an AI model client access Qiniu Cloud storage, intelligent multimedia processing, and live-streaming services through the Model Context Protocol, all from within the model's conversational context. Its storage capabilities cover listing buckets, listing files within a bucket, uploading local files or file content supplied directly, reading file content, and getting a file's download link. Intelligent multimedia capabilities cover image resizing and cutting rounded corners into images. CDN capabilities cover refreshing a cached file by its link and prefetching a file by its link. Live-streaming capabilities are the most extensive: creating a live-streaming space (bucket), creating a live stream within that space, listing live-streaming spaces, listing streams, binding push and pull stream domains, retrieving the push and pull live-stream addresses, and getting live-streaming usage.

Use this when an AI assistant needs to manage Qiniu Cloud storage buckets and files, apply basic image transformations, refresh or prefetch CDN content, or set up and monitor live-streaming spaces and streams, all through natural-language requests instead of the Qiniu console or SDK calls directly. Example interactions shown in the source include listing a bucket's resources, reading a specific file's content, applying a 200-pixel rounded corner to an image, refreshing a CDN link, creating a named live-streaming space and stream, binding push/pull domains, and checking the last hour's live-streaming usage. It requires a Qiniu Cloud account with valid access credentials, so it is not usable standalone without one.

Running the server requires Python 3.12 or later and the uv package manager, installable via Homebrew on Mac, a curl install script on Mac/Linux, or a PowerShell script on Windows. For development, the workflow is cloning the repository, creating and activating a virtual environment with uv venv, installing dependencies with uv pip install -e ., and copying .env.example to .env to configure QINIU_ACCESS_KEY, QINIU_SECRET_KEY, QINIU_REGION_NAME, QINIU_ENDPOINT_URL, and a comma-separated QINIU_BUCKETS list (recommended at up to 20 buckets). The server can be tested with the Model Context Protocol Inspector, or started locally in stdio mode (the default) or SSE mode for web applications via a --transport sse --port flag.

In Cline (as a VS Code extension), the server is configured with uvx as the command and qiniu-mcp-server as the argument, passing the Qiniu credentials as environment variables:

{
  "mcpServers": {
    "qiniu": {
      "command": "uvx",
      "args": [
        "qiniu-mcp-server"
      ],
      "env": {
        "QINIU_ACCESS_KEY": "YOUR_ACCESS_KEY",
        "QINIU_SECRET_KEY": "YOUR_SECRET_KEY",
        "QINIU_REGION_NAME": "YOUR_REGION_NAME",
        "QINIU_ENDPOINT_URL": "YOUR_ENDPOINT_URL",
        "QINIU_BUCKETS": "YOUR_BUCKET_A,YOUR_BUCKET_B"
     },
      "disabled": false
    }
  }
}

For live-streaming-only usage, two lighter authentication modes are supported: QINIU_ACCESS_KEY/QINIU_SECRET_KEY, or a QINIU_LIVE_API_KEY obtained from the Qiniu live-streaming console. Cursor can use the same configuration directly. Claude Desktop users may hit a spawn uvx ENOENT error, resolved by pointing command at the absolute path to the uvx binary (for example /usr/local/bin/uvx) instead of the bare command name.

Developers already using Qiniu Cloud for object storage, image processing, CDN, or live streaming who want to manage those resources conversationally through an MCP-enabled client like Cline, Cursor, or Claude Desktop instead of switching to the Qiniu console.