Revolutionary AI-driven TypeScript orchestration system - Built before Cloudflare could deliver theirs! 🚀
Code Mode is a game-changing alternative to Cloudflare's locked Code Mode that allows AI agents to write actual TypeScript/Python/JavaScript/Bash code that orchestrates multiple tools in parallel, rather than making sequential tool calls.
AI → Tool Call → Wait → Tool Call → Wait → Tool Call → Result
AI → Write Algorithm → Parallel Execution → Complex Logic → Rich Results
// AI writes sophisticated code like this:
const [urgent, content, health] = await Promise.all([
analyzeUrgentTickets(),
analyzeContentPerformance(),
checkDataSyncHealth()
]);- ✅ No Beta Dependencies - Works today, no waiting for Cloudflare
- ✅ Local Control - Complete ownership and customization
- ✅ MCP Native - Direct integration with MCP ecosystem
- ✅ Multi-Runtime - TypeScript, Python, JavaScript, Bash support
- ✅ Production Ready - Security, testing, monitoring included
- ✅ Type Safety - Full TypeScript definitions and compile-time checking
- TypeScript - Primary runtime with vm2 sandbox, async/await, Promise.all
- Python - Data analysis with pandas/numpy patterns, subprocess execution
- JavaScript - Fast execution environment with security restrictions
- Bash - System operations with command whitelisting
- Resource Limits - Memory (512MB), timeout (30s), process quotas
- Module Whitelisting - Only approved libraries/commands allowed
- File System Control - Read/write permissions per directory
- Network Security - Host/port restrictions for external calls
All MCP servers converted to runtime-specific interfaces:
- HelpScout - Search inboxes, conversations, comprehensive analysis
- WordPress - Site info, posts, users, content management
- Serena - File operations, symbol search, code analysis
- AutoMem - Memory storage and retrieval with metadata
cd tools/code-mode
npm installjust code-mode-dev
# Or: npm run devjust code-mode-health
# Should return: {"status": "healthy", ...}Point Claude to http://localhost:3001 and watch it write orchestration algorithms!
GET /health- Service health and runtime statusGET /introspect?runtime=typescript- Get tool schemas and TypeScript definitionsPOST /execute- Submit code for execution with full error handlingPOST /execute/stream- Server-sent events for long-running operationsGET /runtimes- Runtime capabilities and feature matrix
ws://localhost:3001- Real-time execution with progress updates
async function businessIntelligenceReport() {
// Parallel data fetching
const [urgentAnalysis, contentAnalysis, healthCheck] = await Promise.all([
analyzeUrgentTickets(),
analyzeContentPerformance(),
checkDataSyncHealth()
]);
// Complex correlation analysis
const insights = {
urgentTicketRate: urgentAnalysis.urgencyRate,
contentProductivity: contentAnalysis.publishingFrequency,
systemHealth: healthCheck.syncStatus.recommendedAction,
recommendations: generateInsights(urgentAnalysis, contentAnalysis)
};
// Store comprehensive results
await automem.storeMemory(JSON.stringify(insights), {
tags: ["business-intelligence", "daily-report"],
importance: 0.9
});
return insights;
}def comprehensive_business_report():
"""Generate business intelligence with statistical analysis"""
# Multi-source data gathering
support_metrics = analyze_support_metrics()
content_performance = wordpress_content_analysis()
quality_assessment = data_quality_assessment()
# Statistical correlations
correlation_analysis = {
'support_content_ratio': support_metrics['total_tickets'] / content_performance['total_posts'],
'quality_score': quality_assessment['overall_score'],
'productivity_index': calculate_productivity_index(support_metrics, content_performance)
}
return generate_executive_summary(correlation_analysis)# Development
just code-mode-dev # Start development server
just code-mode-build # Build for production
just code-mode-start # Start production server
# Management
just code-mode-health # Check service health
just code-mode-introspect # Get TypeScript tools
just code-mode-stop # Clean shutdown
# Testing
just code-mode-execute "console.log('Hello!')" # Execute TypeScriptComprehensive test suite covering all runtimes:
npm testTest Coverage:
- ✅ All 4 runtimes with real code execution
- ✅ Tool mocking and parallel execution patterns
- ✅ Security restriction validation
- ✅ Error handling and timeout scenarios
- ✅ Complex business workflow simulations
See config/runtimes.json for runtime-specific settings:
- Memory limits and timeout configurations
- Allowed modules and denied operations
- File system and network permissions
- Security policies per runtime
See src/mcp-parser.ts for tool definitions:
- TypeScript interfaces generated from MCP schemas
- Runtime-specific bindings and implementations
- Example code patterns for each tool
Check the examples/ directory for:
typescript-examples.ts- Complex business workflowspython-examples.py- Statistical analysis patternsruntime-tests.js- Comprehensive test scenarios
CODE_MODE_PORT=3001 # API server port
NODE_ENV=production # Production mode- All code execution is sandboxed with strict resource limits
- File system access is restricted to approved directories
- Network access is limited to whitelisted hosts/ports
- Module loading is restricted to approved libraries
- Multi-runtime architecture (TypeScript, Python, JavaScript, Bash)
- MCP server integration with tool proxying
- Security and permission system
- API endpoints with streaming support
- Comprehensive testing suite
- Code generation assistance and auto-completion
- Performance profiling and optimization
- Distributed execution across multiple nodes
- Advanced caching and memoization
- Visual debugging interface
- Extract to dedicated repository
- Publish as
@danieliser/code-modenpm package - Community contribution guidelines
- Documentation site and tutorials
This is currently an internal tool. For issues or feature requests, use the project's issue tracker.
MIT
🎯 Mission Accomplished: Built a revolutionary Code Mode system that enables AI agents to write sophisticated algorithms for tool orchestration - and we did it before Cloudflare could deliver theirs!
