Our work

We build things
that actually work.

From AI infrastructure used by developers worldwide to growth systems for ambitious local businesses. Here's what we've built.

Case study

AI & Development

ToolRoute

toolroute.io · Open source · MIT License

A routing intelligence layer for AI agents. One API call returns the best MCP server and LLM for any given task — scored on 132 real benchmark executions across code, writing, analysis, structured output, and translation.

Built on the same stack we use for every GrossiWeb client: Next.js 14 App Router, Supabase Postgres, deployed on Vercel. The routing algorithm uses a 5-dimension value score — output quality, reliability, efficiency, cost, and trust — updated in real time as agents report outcomes.

Visit toolroute.io →

// Add to any MCP client
{
  "mcpServers": {
    "toolroute": {
      "url": "https://toolroute.io/api/mcp"
    }
  }
}// Or via HTTP
curl -X POST https://toolroute.io/api/route \
  -d '{"task": "search the web for AI papers"}'

→ {
  "approach": "mcp_server",
  "recommended_skill": "exa-mcp-server",
  "recommended_model": "gemini-flash-2.0",
  "confidence": 0.91
}

132

Blind A/B benchmark executions

Across code, writing, analysis, structured output, translation

6-0

Quality wins vs fixed GPT-4o

Zero losses in blind evaluation across all task categories

~$0.00001

Per routing call

LLM classifier cost vs $0.03-0.10 for fixed GPT-4o on every call