Agent-Mon

Agent-Mon is for non technical users who cant spawn agents or use Claude Code. Agent-Mon reimagines AI agent orchestration as a pixel-art RPG — no terminal, no IDE, no chat box. Non-technical users walk an avatar through a living neighbourhood where sub-agents hatch from Pokéballs, march to their workstations, and return with findings on parchment scrolls. You accept or retry with A/B — the most battle-tested two-button interface ever made.
The interface itself is the innovation. Districts map to product phases (Explore → Design → Build). Proximity to a character triggers a Haiku-powered status quip instantly; pressing Space fires a Sonnet deep-analysis returning four structured findings; Deploy synthesises everything into a shipped landing page. Claude’s structured JSON output is the world content — the game cannot exist without the intelligence.
Stability comes from radical simplicity: a single self-contained HTML client, three clean Node.js endpoints, Redis for audit logging, and productContext threaded through every API call so agents never lose the thread.
Stack: Claude Sonnet & Haiku (Anthropic SDK), Node.js, Vercel serverless, Redis (Upstash), HTML5 Canvas, vanilla JS, CSS with Press Start 2P typeface — no bundler, no framework, no build step.
Building software should feel like play, not configuration.

NA