serve Command¶
Start the MCP Standards Server for Model Context Protocol integration.
Synopsis¶
Description¶
The serve command starts the MCP Standards Server, which provides a Model Context Protocol interface for LLMs and development tools to interact with standards programmatically. This enables real-time standard queries, validation, and code generation assistance.
Options¶
--host <address>¶
Host address to bind to (default: localhost).
--port <number>¶
Port to listen on (default: 3000).
--stdio¶
Run in stdio mode for direct tool integration.
--socket <path>¶
Use Unix domain socket instead of TCP.
--workers <n>¶
Number of worker processes (default: CPU count).
--log-level <level>¶
Set logging level (debug, info, warning, error).
--auth <type>¶
Enable authentication (none, token, oauth).
--tls-cert <file>¶
TLS certificate file for HTTPS.
--daemon¶
Run as background daemon.
Examples¶
Basic Server Start¶
Output:
Starting MCP Standards Server v1.0.0
Loading standards from cache...
✓ Loaded 45 standards files
✓ Initialized rule engine
✓ Search index ready
✓ Token optimizer configured (gpt-4, 8000 tokens)
Server listening on http://localhost:3000
Available MCP tools:
- get_applicable_standards
- validate_code
- search_standards
- get_standard_content
- check_compliance
Press Ctrl+C to stop
Production Server¶
mcp-standards serve \
--host 0.0.0.0 \
--port 443 \
--workers 8 \
--auth token \
--tls-cert /etc/ssl/certs/mcp.crt \
--tls-key /etc/ssl/private/mcp.key \
--log-level info \
--daemon
Stdio Mode for Tool Integration¶
When running in stdio mode, the server communicates via standard input/output:
{"jsonrpc": "2.0", "method": "initialize", "params": {}, "id": 1}
{"jsonrpc": "2.0", "result": {"capabilities": {...}}, "id": 1}
Unix Socket Mode¶
MCP Tools Available¶
get_applicable_standards¶
Get standards based on project context:
{
"tool": "get_applicable_standards",
"arguments": {
"context": {
"project_type": "web-application",
"frameworks": ["react", "nextjs"],
"languages": ["typescript"],
"requirements": ["accessibility", "performance"]
},
"include_resolution_details": true
}
}
validate_code¶
Validate code against standards:
{
"tool": "validate_code",
"arguments": {
"code": "const Button = ({onClick}) => <button onClick={onClick}>Click</button>",
"language": "javascript",
"standards": ["react-18-patterns", "wcag-2.2-accessibility"]
}
}
search_standards¶
Search standards using natural language:
{
"tool": "search_standards",
"arguments": {
"query": "How to implement secure authentication in Node.js?",
"limit": 5,
"include_content": true
}
}
get_standard_content¶
Retrieve specific standard content:
{
"tool": "get_standard_content",
"arguments": {
"standard_id": "react-18-patterns",
"format": "condensed",
"token_budget": 2000
}
}
Configuration¶
Server configuration in .mcp-standards.yaml:
server:
host: localhost
port: 3000
workers: auto # or specific number
# Logging
log_level: info
log_file: /var/log/mcp-standards.log
# Authentication
auth:
type: token # none, token, oauth
token_file: /etc/mcp-standards/tokens.json
oauth:
provider: github
client_id: xxx
client_secret: xxx
# TLS/SSL
tls:
enabled: false
cert: /path/to/cert.pem
key: /path/to/key.pem
ca: /path/to/ca.pem
# Performance
cache:
enabled: true
ttl: 3600
max_size: 1000
# Token optimization
token_optimization:
enabled: true
default_budget: 8000
model_type: gpt-4
# Rate limiting
rate_limit:
enabled: true
requests_per_minute: 60
burst: 100
Authentication¶
Token Authentication¶
# Generate tokens
mcp-standards serve --generate-token --name "vscode-client"
Token generated: mcp_token_a1b2c3d4e5f6
# Start server with token auth
mcp-standards serve --auth token
Client usage:
curl -H "Authorization: Bearer mcp_token_a1b2c3d4e5f6" \
http://localhost:3000/tools/get_applicable_standards
OAuth Authentication¶
# Configure OAuth
mcp-standards config --set server.auth.type oauth
mcp-standards config --set server.auth.oauth.provider github
# Start server
mcp-standards serve --auth oauth
Monitoring¶
Health Check Endpoint¶
Response:
{
"status": "healthy",
"version": "1.0.0",
"uptime": 3600,
"standards_loaded": 45,
"cache_size": 3.42,
"workers": 4,
"memory_usage_mb": 128.5,
"last_sync": "2025-07-08T10:15:30Z"
}
Metrics Endpoint¶
Prometheus format output:
# HELP mcp_requests_total Total number of MCP requests
# TYPE mcp_requests_total counter
mcp_requests_total{tool="get_applicable_standards"} 1234
# HELP mcp_request_duration_seconds Request duration in seconds
# TYPE mcp_request_duration_seconds histogram
mcp_request_duration_seconds_bucket{le="0.1"} 950
Logging¶
# View logs
tail -f /var/log/mcp-standards.log
# Log format
2025-07-08 10:30:45 INFO: Request received: get_applicable_standards
2025-07-08 10:30:45 DEBUG: Context: {project_type: "web", frameworks: ["react"]}
2025-07-08 10:30:45 INFO: Found 7 applicable standards
2025-07-08 10:30:45 INFO: Response sent in 45ms
Integration Examples¶
VS Code Extension¶
// .vscode/settings.json
{
"mcp-standards.server.url": "http://localhost:3000",
"mcp-standards.server.token": "mcp_token_xxx",
"mcp-standards.validation.onSave": true,
"mcp-standards.suggestions.enabled": true
}
Claude Desktop Configuration¶
// claude_desktop_config.json
{
"mcpServers": {
"standards": {
"command": "mcp-standards",
"args": ["serve", "--stdio"],
"env": {
"MCP_STANDARDS_CONFIG": "/home/user/.config/mcp-standards/config.yaml"
}
}
}
}
Docker Deployment¶
FROM python:3.11-slim
RUN pip install mcp-standards-server
COPY config.yaml /etc/mcp-standards/
EXPOSE 3000
CMD ["mcp-standards", "serve", \
"--host", "0.0.0.0", \
"--port", "3000", \
"--config", "/etc/mcp-standards/config.yaml"]
Systemd Service¶
# /etc/systemd/system/mcp-standards.service
[Unit]
Description=MCP Standards Server
After=network.target
[Service]
Type=simple
User=mcp-standards
Group=mcp-standards
WorkingDirectory=/var/lib/mcp-standards
ExecStart=/usr/local/bin/mcp-standards serve --daemon
Restart=always
RestartSec=5
[Install]
WantedBy=multi-user.target
Nginx Reverse Proxy¶
server {
listen 443 ssl http2;
server_name standards.example.com;
ssl_certificate /etc/ssl/certs/example.crt;
ssl_certificate_key /etc/ssl/private/example.key;
location / {
proxy_pass http://localhost:3000;
proxy_http_version 1.0.0
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection 'upgrade';
proxy_set_header Host $host;
proxy_cache_bypass $http_upgrade;
# MCP specific headers
proxy_set_header X-MCP-Client $remote_addr;
proxy_read_timeout 300s;
}
}
Performance Tuning¶
Worker Configuration¶
# Auto-detect optimal workers
mcp-standards serve --workers auto
# Manual configuration for high load
mcp-standards serve --workers 16 --log-level warning
Memory Optimization¶
# config.yaml
server:
memory:
max_heap_mb: 2048
gc_interval: 300 # seconds
cache_strategy: lru
preload_standards: true
Connection Pooling¶
Troubleshooting¶
Debug Mode¶
Common Issues¶
-
Port Already in Use
-
Standards Not Loading
-
Authentication Failures