Supergateway

Supergateway is an MCP server that acts as a transport proxy, converting Model Context Protocol servers between standard input/output (stdio) and network transports such as Server-Sent Events (SSE), WebSockets (WS), and Streamable HTTP. Developers, system administrators, and AI engineers use Supergateway to expose local command-line MCP servers over HTTP or WebSockets for remote consumption, or conversely, to bridge remote HTTP and SSE MCP servers into local stdio-only MCP clients. The tool allows users to run stdio-based servers as web services, attach authentication headers or OAuth bearer tokens to remote requests, configure CORS policies, define custom endpoints, and maintain stateful sessions. Supergateway connects developer machines, containerized workloads, and remote servers directly to MCP-compliant environments like Claude Desktop, Cursor, and web-based AI tooling.

Category: AI & LLM Tooling

Tags: gateway, proxy, sse, transport, websockets

Visit Supergateway

How to install and configure Supergateway

Follow these steps to run Supergateway or bridge remote MCP servers into your client: 1. Run Supergateway directly using npx or Docker without local package installation: bash npx -y supergateway --stdio "uvx mcp-server-git" 2. To expose an SSE server into Claude Desktop or Cursor via stdio, edit your client configuration file (claude_desktop_config.json or Cursor MCP settings): json { "mcpServers": { "supergateway": { "command": "npx", "args": [ "-y", "supergateway", "--sse", "https://mcp-server.example.com" ] } } } 3. If your remote server requires authorization, add authentication flags such as --oauth2Bearer "your-token" or --header "Header-Name: value" directly to the args array.

What you can do with Supergateway

• Exposing local stdio MCP servers over Server-Sent Events or WebSockets to allow remote web clients and inspectors to interface with them. - Bridging remote SSE or Streamable HTTP MCP endpoints into desktop clients like Claude Desktop or Cursor that only support stdio execution. - Running local MCP file-system or command-line utilities inside Docker containers while mapping communication to HTTP ports. - Tunneling local stdio MCP processes over internet connections using tools like ngrok to debug agent workflows remotely. - Providing stateful Streamable HTTP endpoints with defined session timeouts for modern MCP application deployments.

Key facts

• https://github.com/supercorp-ai/supergateway
• AI & LLM Tooling
• gateway, proxy, sse, transport, websockets

Part of MCP Servers

What is Supergateway?

Supergateway is an MCP proxy tool that translates Model Context Protocol communication between stdio and network transports. It allows developers to run standard command-line MCP servers over Server-Sent Events, WebSockets, or Streamable HTTP, and conversely exposes remote HTTP and SSE endpoints locally via stdio.

How do I install Supergateway?

You can run Supergateway without a permanent installation using npx with the command npx -y supergateway followed by your transport flags. Alternatively, official Docker images are available under the supercorp/supergateway repository on Docker Hub and GitHub Container Registry, including preconfigured images with uvx and Deno.

Which MCP clients work with Supergateway?

Supergateway works with any MCP-compliant client. When running in SSE-to-stdio or Streamable HTTP-to-stdio modes, it connects remote services to local clients such as Claude Desktop and Cursor. In stdio-to-SSE mode, it supports web clients like the MCP Inspector.

How does Supergateway handle authentication headers?

Supergateway accepts custom HTTP headers using the --header flag, which can be repeated for multiple headers. For OAuth access tokens, the dedicated --oauth2Bearer flag adds standard Bearer authorization headers, which avoids formatting errors in environments like Cursor that struggle with spaces in arguments.