MCP SSE Launcher
A Python management system for MCP servers that wraps stdio-based servers as SSE endpoints and includes a web inspector for testing.
MCP SSE Launcher
A Python-based management system for Model Context Protocol (MCP) servers that automatically wraps stdio-based servers with Supergateway to expose them as SSE endpoints, plus provides a web-based inspector for testing.
Key Features
- Auto-update - keeps MCP servers up-to-date automatically
- Background monitoring - restarts crashed servers automatically
- Web inspector - debug and test MCP servers via browser
- SSE conversion - wraps stdio servers for HTTP/SSE access
- AI assistant integration - works with Cursor, Claude Desktop, etc.
Quick Start
# 1. Install dependencies
pip install -r requirements.txt
# 2. Set up environment (copy .env.example to .env and add API keys)
cp .env.example .env
# 3. Start all MCP servers (with auto-restart)
make run-servers
# 4. Start web inspector for debugging
make run-inspector
# 5. View inspector at the URL shown in the output
Common Use Cases
Debug a specific MCP server
# Start inspector (will show URL with auth token)
make run-inspector
# Visit the URL shown in output to test MCP servers
Add a new MCP server
# Add GitHub MCP server
python mcp_servers.py add \
--name github \
--cmd npx \
--args "-y @modelcontextprotocol/server-github" \
--env GITHUB_TOKEN=your_token_here \
--port 8094
Use with Cursor AI
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"filesystem": { "url": "http://localhost:8090/sse" },
"github": { "url": "http://localhost:8094/sse" },
"brave-search": { "url": "http://localhost:8096/sse" }
}
}
Restart a problematic server
# Kill conflicts and restart specific server
make restart-server SERVER=github
Run without auto-restart (for development)
make run-servers KEEP_ALIVE=0
What Happens Automatically
- Updates: npm packages and git repos are checked/updated before starting
- Port conflicts: conflicting processes are detected and can be killed
- Process monitoring: crashed servers restart up to 3 times
- Logging: each server logs to
/tmp/<server_name>.log - Cleanup: all processes are properly terminated on exit
Available MCP Servers
The default config includes popular servers like:
- filesystem (8090) - local file access
- github (8094) - GitHub repository interaction
- brave-search (8096) - web search
- fetch (8093) - HTTP requests
- youtube (8092) - YouTube search/info
- memory (8100) - persistent memory
- puppeteer (8101) - web automation
Troubleshooting
# Check what's using server ports
make check-server-ports
# Kill conflicting processes on server ports
make kill-server-conflicts
# Stop servers (inspector stops with Ctrl+C)
make stop-servers
# View server logs
tail -f /tmp/filesystem.log
Configuration
Servers are defined in mcp_config.json. Environment variables use ${VAR} syntax and are loaded from .env.
Example server config:
{
"name": "github",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": {"GITHUB_TOKEN": "${GITHUB_TOKEN}"},
"port": 8094,
"server_type": "stdio"
}
Related Servers
Ant Design Components
Provides Ant Design component documentation to large language models (LLMs), allowing them to explore and understand the components.
HeyBeauty
Perform virtual try-ons using the HeyBeauty API.
MCP API Bridge
A server that bridges Google Sheets, Azure AI, and MQTT APIs.
MCP Tool Builder
An MCP server that allows LLMs to dynamically create new tools through MCP clients like Claude Desktop.
DocsetMCP
A server for accessing Dash-style documentation sets locally. Requires a local Dash installation.
MCP OpenAPI Connector
Connect to any OpenAPI-based API with built-in OAuth2 authentication management.
Remote MCP Server (Authless)
A remote MCP server deployable on Cloudflare Workers without authentication.
Shell MCP
Securely execute shell commands with whitelisting, resource limits, and timeout controls for LLMs.
eBPF MCP
A secure MCP server for eBPF, designed for AI integration, kernel introspection, and automation.
Vibetest Use
A browser-agent QA swarm with an MCP interface for testing AI-generated websites.