Show HN: Middleware for running autonomous AI coding agents in sandboxes

We've built what we believe is the missing piece in AI development: an agent-agnostic middleware infrastructure that lets AI coding assistants run in parallel sandboxed environments. Key technical features: - Dynamic workspace provisioning with parallel sandbox environments - Connect to the workspace with VS Code - Full system interaction capabilities (shell, Git, LSP) - Resource-optimized cloud infrastructure - Enterprise-grade security and access controls Why we put together this proof-of-concept: Current AI coding agents are limited by working in single files or environments. We have a vision to enable agents to: - Spin up multiple sandboxes to test different solutions simultaneously - Access full development environments, not just single files - Run real-time tests - Scale compute resources efficiently Our implementation addresses the core issues outlined in our CEO's recent analysis of AI coding infrastructure (https://go.daytona.io/ai-coding https://go.daytona.io/ai-coding). This is a proof of concept demonstrating how AI agents can work with proper infrastructure support. We're looking for feedback from the HN community, especially: - Developers building AI coding assistants - Teams working on development environment tooling - Anyone interested in AI agent infrastructure standards The code is open source and we welcome contributions. Happy to answer any technical questions! --- Built at Daytona - we make dev environment tools