CLU
architecture.md
// architecture.md

Architecture

Five layers. One complete ecosystem. Each layer gives you more power — from natural interfaces to platform modules with vault, jobs, graphs and memory.

80%Cost reduction
70%Faster
ROI Year 1
90%Automation rate

Five layers. One complete ecosystem.

Your InterfaceTrigger
Chat · API REST · MCP · Webhook · Cron
SmartAgentsReasoning
Multi-step flows with autonomous reasoning and adaptation
Skills<50ms
Pre-defined patterns. Zero tokens. Response <50ms
ConnectorsEcosystem
Google Sheets · SendGrid · HubSpot · SAP · WhatsApp · Slack
Platform ModulesCore
Vault (AES-256) · Jobs · Graph · Memory · RAG

Agent Network: CLU as Hub

CLU connects bidirectionally with external AI agents via MCP. Claude, ChatGPT, custom agents and your own bots can invoke CLU tools — and CLU can invoke theirs.

Bidirectional MCP Flow

Any external agent can call CLU, and CLU can call external agents. All via standard JSON-RPC 2.0.

→ External agent invokes CLU
Claude Desktop
MCP JSON-RPC
CLU Router
SmartAgent
Connectors
Response
← CLU invokes external agent
CLU Agent
MCP Client
External Agent
Process
Result → CLU
↔ Multi-Agent Orchestration
Agent A
CLU Hub
Agent B
CLU Hub
Agent C

CLU acts as central hub: each agent talks to CLU, and CLU coordinates the conversation between all of them. No agent needs to know the others directly.