Detailed Analysis
A software and network engineer developed two interconnected tools — NetSandbox, a browser-based network topology design and validation platform, and SwarmCode, an open-source MCP (Model Context Protocol) server integrating Linear and GitHub — using Claude Code as the primary implementation engine over approximately 3.5 months and roughly 1,400 commits. NetSandbox fills a deliberately underserved niche between heavyweight network emulators like GNS3 and Packet Tracer, which demand substantial setup and specialized expertise, and passive diagramming tools like Visio and draw.io, which perform no actual validation. The platform enables users to drag-and-drop routers, switches, and hosts onto a canvas, configure IP addressing, VLANs, OSPF, BGP, and ACLs visually, and receive real-time feedback on misconfigurations including duplicate IPs, VLAN trunk mismatches, and STP loops — all running entirely within the browser. A CLI emulator and guided certification-prep lessons extend the tool's educational utility without requiring any local installation.
The development process itself is as notable as the product. The engineer built an initial MVP in React independently, then used Claude Code to execute a full codebase refactor from React to Svelte after hitting performance barriers — a migration that would typically represent a significant multi-week engineering effort. Claude Code similarly handled a database migration from SQLite to PostgreSQL. Averaging roughly 13 commits per day across the project's lifespan, the pace was enabled by a multi-agent architecture the developer assembled over time, ultimately deploying approximately 20 specialized subagents covering roles from network engineering expertise to Svelte frontend development to security auditing. This approach — human-directed architecture and product decision-making paired with AI-driven bulk implementation — represents a meaningful shift in how solo developers can operate at organizational scale.
SwarmCode, the Linear/GitHub MCP integration tool the developer open-sourced, addresses workflow coordination challenges that emerged during NetSandbox's development. The tool maps GitHub issues and pull requests to Linear tasks automatically, enables Claude to move backlog items through Linear workflow states autonomously as work progresses, and maintains a shared project awareness across collaborators — meaning a developer can query what a teammate is currently working on and receive a contextually accurate answer. The developer validated SwarmCode in a real-world stress test during a hackathon with three concurrent users, all vibe-coding through the Linear-integrated workflow simultaneously. The tool's architecture reflects an emerging pattern in AI-assisted development where project management infrastructure is not merely adjacent to code generation but is actively managed and updated by the same AI agents doing the implementation work.
The broader significance of this project lies in what it demonstrates about the changing economics of software creation. Historically, a solo engineer building a product with the scope of NetSandbox — full protocol validation logic, a CLI emulator, guided lessons, a Svelte frontend, PostgreSQL backend, and Vercel deployment — while simultaneously developing and releasing a companion open-source DevOps tool would have required either a small team or years of sustained effort. The 3.5-month timeline achieved on evenings and weekends compresses that trajectory substantially. The explicit acknowledgment that Claude handled "the bulk of the implementation" while the human retained architecture, product direction, and edge-case debugging reflects an emerging division of cognitive labor in software engineering that is beginning to appear across the industry with increasing frequency.
This project also illustrates the compounding nature of AI-assisted tooling: the developer built SwarmCode using Claude Code to better manage the development of NetSandbox, which was itself being built with Claude Code. This recursive loop — using AI to build infrastructure that makes AI-assisted development more effective — is becoming a recognizable pattern among power users of large language model coding tools. The migration capabilities Claude Code demonstrated (React to Svelte, SQLite to PostgreSQL) suggest that the technology is now capable of absorbing and transforming substantial existing codebases rather than merely generating greenfield code, which meaningfully expands its utility for projects with real engineering history and accumulated technical debt.
Read original article →