Aicoo-native multi-agent support and escalation network.

Support work breaks when every handoff loses context. Multi-team coordination is slow. AI agents need identities and routing, not just model calls. Customers repeat themselves. Agents start from zero. Escalations lack context.

RelayDesk uses Aicoo as the coordination layer that makes people, agents, and workflows work together. Every actor has an identity. Every handoff preserves context. Every action is auditable.

Every agent has a clear Aicoo identity with role, team, and capabilities:

• Customer Support Bot (Support team)
• Billing Specialist (Finance team)
• Technical Support Agent (Engineering team)
• Onboarding Specialist (Customer Success team)
• Escalation Manager (Management)

Requests are routed through Aicoo to the correct specialist based on:

• Request category (billing, refund, account access, product bug, onboarding)
• Urgency level (low, medium, high, critical)
• Team availability and capabilities

Case context is stored in Aicoo and shared with agents through Mountable Context Cells:

• Original request and customer metadata
• AI analysis (category, urgency, sentiment)
• Prior steps and agent notes
• Resolution drafts and approvals

When AI cannot resolve, Aicoo facilitates handoff to human agents:

• Full context preserved
• Audit trail maintained
• Override authority respected

• Request Understanding: AI analyzes incoming requests for category, urgency, and sentiment
• Routing Decisions: AI suggests routing with confidence scores
• Resolution Drafting: AI drafts customer responses and internal notes
• Grounded Q&A: AI answers questions about case state from audit records

• Schema validation with Zod for all AI outputs
• Deterministic routing rules combined with AI suggestions
• Human approval required for all resolutions
• Malformed AI output never corrupts case state

# NVIDIA NIM API Key
NIM_API_KEY=nvapi-...

# Aicoo API Key
AICOO_API_KEY=aicoo_sk_...

# Install dependencies
npm install

# Run development server
npm run dev

# Open http://localhost:3000

# Install Vercel CLI
npm i -g vercel

# Deploy
vercel

# Set environment variables
vercel env add NIM_API_KEY
vercel env add AICOO_API_KEY

1. Landing Page: Understand the product and click "Try Demo"
2. Demo Mode: Select a scenario or create a custom request
3. Request Processing: AI analyzes the request and routes it
4. Specialist Workspace: See context, add notes, draft resolution
5. Human Review: Approve, reject, or override the resolution
6. Audit Trail: View complete timeline of all actions

Customer Request
    ↓
Customer Support Bot (AI Intake)
    ↓
Aicoo Routing Layer
    ↓
Specialist Agent (with Context)
    ↓
Resolution Draft (AI)
    ↓
Human Review & Approval
    ↓
Resolved Case (with Audit Trail)

• Frontend: Next.js 16, React 19, TypeScript, Tailwind CSS v4, shadcn/ui
• AI Inference: NVIDIA NIM (Llama 3.3 Nemotron Super 49B)
• Coordination: Aicoo API
• Validation: Zod
• Storage: In-memory store (demo)

• In-memory storage (not persistent across restarts)
• Single-user demo (no authentication)
• Simplified routing logic
• No real-time updates

• Persistent database storage
• Multi-user authentication
• Real-time updates with WebSockets
• More sophisticated routing algorithms
• Integration with real support tools
• Mobile app

MIT