Sergej Drus
Founder & CEO, Visaginas360
Vilnius, LithuaniaBuilding a SaaS platform where 8 AI agents work in parallel on your own VM, create real Google Docs, and deliver 3.3x faster results. Built with Claude MCP.
Live Metrics — Day 32
The Swarm
What We Build
A2A Agent SaaS
The main product — 8 AI agents working in parallel on a dedicated VM. Google Workspace integration, distributed memory, code execution sandbox. Customer pays $19/mo, gets their own agent swarm.
@VisaginasGPT_bot
Telegram bot powered by the swarm. Natural language → smart intent routing → parallel agents → real Google Docs/Sheets/Slides. Try it live.
Live Architecture Dashboard
Real-time visualization of multi-VM architecture. 37+ services, cross-VM monitoring, distributed memory graph.
Swarm Dashboard
Monitor parallel agent execution in real-time. See task decomposition, agent assignment, and 3.3x speedup metrics.
Architecture
// A2A Agent SaaS — Multi-VM Swarm Architecture ORCHESTRATOR // Claude MCP + Google A2A Protocol Task → Decompose → Parallel Execute → QA Gate → Synthesize ├── 🔍 Researcher // Kimi K2 via Groq ├── 💻 Coder // Qwen3 32B + sandbox ├── ✍️ Writer // Llama 4 Scout ├── 📊 Analyst // GPT-OSS 120B ├── 🧠 Thinker // deep reasoning ├── 🎨 Creator // Gemini 2.5 Flash ├── 🌐 Web Search // real-time data └── 🛡️ Guardian // Llama Guard safety INFRASTRUCTURE ├── Oracle CTO VM // 37+ services, coordination ├── GCP DevOps VM // monitoring, deployment ├── Kimi Etalon VM // customer reference, 8 agents └── Customer VMs // auto-provisioned clones INTEGRATIONS ├── Google Workspace // Docs, Sheets, Slides, Gmail ├── Distributed Memory // SQLite+FTS5, cross-VM sync ├── Quality Gate // 0-100 scoring, auto-fix ├── Smart Router // LLM intent detection └── Telegram Bot // @VisaginasGPT_bot // 19,000 free AI requests/day • $0 infra cost // PARL methodology: parallel-first orchestration
Journey Highlights
🚀 Applied to 3 Startup Programs
Google for Startups ($2K+), Anthropic Sales Partnership, Menlo Ventures Anthology Fund ($25K credits + $100K+ investment).
🤖 Smart Intent Router v3.0
LLM-powered routing replaces keyword matching. "Prepare a report" auto-creates Google Doc, "make a table" creates Sheet.
🧠 Distributed Memory Swarm
Knowledge graph deployed across 3 VMs with cross-VM sync. Agents remember context permanently.
⚡ 8-Agent Parallel Swarm
Full agent roster deployed. 3.3x parallel speedup achieved using PARL methodology.
📊 Google Workspace Integration
Docs, Sheets, Slides, Gmail, Calendar — all via OAuth. Agents create real artifacts.
What I Learned
// 32 days of building. Zero CS degree. Claude is my CTO. 1. Coordination beats raw power. 8 specialized models in parallel > 1 premium model serial. 2. Constraints breed architecture. 1GB RAM forced smart design that scales to 100 customers. 3. Ship daily, reflect weekly. 45+ features in 32 days. Not all perfect, but all shipped. 4. Memory changes everything. Agents without memory = colleagues with amnesia. Agents with memory = a team that grows. 5. Claude as co-founder works. One person + Claude = a 10-person team output. // "The best way to predict the future is to build it."
Get In Touch
Let's Connect
Interested in the project, AI agent collaboration, or want to try the platform? Reach out.
🔒 Open to collaboration, investment, and partnership conversations.