Every manufacturing flow got digitized.
Except the one coming out the back door.
Factories run SCADA, MES, and predictive maintenance on every production line — then sell 50 tonnes of monthly scrap to whichever dealer called first, at 30–40% below fair value. That asymmetry costs Indian manufacturing an estimated ₹25,000 crore every year.
Six autonomous agents.
One event-driven MCP nervous system.
Each agent is a standalone MCP server. They communicate through Supabase Realtime — when the Intake Agent writes a listing, the Verification Agent wakes up automatically. No polling. No orchestrator bottleneck.
Polls ERP disposal queues, accepts photo uploads, and answers multilingual voice calls from MSME factories. Sellers set their own price.
Gemini Vision grades material health (A/B/C), classifies downstream uses, benchmarks price, and scores seller trust without human intervention.
Natural-language BOM decomposition plus cluster analysis that recommends the best industrial zone to source from based on density, price, and trust.
Solves the many-to-many allocation problem — combines sellers to fill one buyer's BOM at minimum cost plus transport.
Google Maps MCP integration for routes, freight estimates, and transporter matching when a buyer requests delivery assistance.
Reads production schedules and forecasts waste before it exists — then pre-notifies matched buyers via WhatsApp. Predictive supply.
Connect any MCP client in 60 seconds.
Relink speaks the Model Context Protocol over streamable HTTP. Claude, ChatGPT, NitroStudio, or your own agent framework can discover and call all 22 tools with no SDK lock-in.
curl https://relink-6a64ea0e-aravindco-amrita-university-coimbatore.app.nitrocloud.ai/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/list"
}'curl https://relink-6a64ea0e-aravindco-amrita-university-coimbatore.app.nitrocloud.ai/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "search_materials",
"arguments": {
"material_type": "aluminum_scrap",
"max_price_per_kg": 150,
"max_radius_km": 100
}
}
}'{
"mcpServers": {
"relink": {
"url": "https://relink-6a64ea0e-aravindco-amrita-university-coimbatore.app.nitrocloud.ai/mcp",
"transport": "http"
}
}
}All 22 live tools
Built for the factory floor, not the slide deck.
Auto Stamping Plant, Pune
Uploads a photo of 500 kg Al-6061 stamping scrap. Gemini Vision grades it B, benchmarks at ₹148/kg. Manager quotes ₹140/kg. Listed in 40 seconds with a GST-verified trust badge.
Die-Casting Unit, Chakan
Types “need 500 kg aluminum scrap for die-casting, budget ₹145/kg”. Sourcing Agent recommends the Talegaon zone — 7 sellers, average ₹132/kg, Grade A. Buyer taps “Get Seller Contact” and calls directly.
Fabrication MSME, Coimbatore
No ERP, no computer. Calls in, speaks in Tamil: “200 kg mild steel offcuts, want ₹28/kg.” The voice agent transcribes, structures, and lists it — the digital divide closed with a phone call.
Prediction Engine, Every Factory
The production schedule shows a 10,000-panel batch run Thursday. The Prediction Agent knows that means roughly 500 kg scrap by Friday — and pre-notifies three matched buyers on WhatsApp before the waste exists.
Sellers set the price. AI keeps everyone honest.
Every listing shows the seller's own quoted price next to the independent AI benchmark — protecting uninformed sellers from underpricing and giving buyers a fair reference point.