VetoCore
Multi-Agent AI OrchestrationDescribe software and a roster of specialist AI agents (PM → Database → Backend/Frontend → Security → QA) builds it through a phased pipeline, streaming their work into a live console. Web, Windows desktop, and an installable PWA on one account.
Agent Pipeline
- Router briefs PM, then Database / Backend / Frontend / Security / QA
- Phased generation with live console streaming (SSE)
- QA-rollback guardrail before shipping generated apps
Platform & Auth
- Authenticated LLM proxy (Groq / OpenRouter / Gemini) with per-token metering
- JWT + argon2 auth; AES-256-GCM encrypted provider keys
- Role-based user + admin dashboards; MongoDB persistence
Clients
- SolidJS web app with in-browser live preview
- Electron Windows desktop build
- Installable PWA on the same account
- SolidJS
- Electron
- Hono
- MongoDB
- TypeScript
- JWT · AES-GCM
- SSE
- PWA
→ Router Briefing the team — PM engaged.
09:24:01 [PM] Acknowledged, Commander. Delegating.
09:24:07 [Database] Schema + migrations ready.
09:24:11 [Security] No exposed secrets — approved.
09:24:15 [Frontend] Writing index.html▍