EdgeOnto is an AI Agent SaaS platform for manufacturing, deeply integrating factory systems, capabilities, and knowledge with large language models through a four-layer architecture — MCP Hub non-invasively connects all business systems, Skill Engine encapsulates pluggable manufacturing expertise, Knowledge Engine preserves proprietary factory knowledge, and Agent Engine provides LLM-driven intelligent orchestration linking all three into executable business workflows.
EdgeOnto is not a generic AI chatbot, but an operations hub rooted in the manufacturing shop floor — driving cross-system collaboration with natural language, making knowledge searchable, capabilities reusable, and decisions traceable.
Unify ERP, MES, WMS, and IoT connectivity via Model Context Protocol — no system modifications required
Shop floor supervisors can query OEE, create work orders, and analyze quality in natural language — no system training required
OEE analysis, smart scheduling, quality traceability and other manufacturing capabilities delivered as standardized Skills — enable on demand
Start with lightweight scenarios like knowledge Q&A and progressively expand to autonomous execution — validate AI value with low risk
EdgeOnto decomposes the AI Agent infrastructure into three independent yet collaborative engines — corresponding to Arms & Legs, Expertise, and Memory — enabling each layer to iterate independently and evolve in combination.
Non-invasive connection to all systems — Based on the Model Context Protocol (MCP) standard, unify heterogeneous systems such as ERP, MES, WMS, and IoT so that AI Agents gain the ability to read data and execute operations.
Encapsulating pluggable manufacturing expertise — Package manufacturing domain knowledge such as OEE analysis, smart scheduling, and quality traceability into standardized Skill modules that Agents invoke on demand, decoupling capabilities from the platform.
Preserving all proprietary factory knowledge — Organize factory knowledge including SOPs, equipment manuals, BOMs, and historical fault records into a semantically searchable enterprise brain through RAG vector retrieval and knowledge graph technologies.
Build factory knowledge — process documents, equipment manuals, quality standards, fault cases — into a semantically searchable AI knowledge base. Workers get precise answers in natural language.
Monitor production data in real time, automatically identify abnormal fluctuations in key metrics such as output, quality, and energy consumption, and generate actionable corrective plans through LLM root cause analysis.
An interactive AI assistant in digital human form, supporting both voice and text input, enabling frontline workers to access information without navigating complex systems.
Based on multi-source data fusion and AI reasoning, provide management with multi-scenario comparison and risk assessment for key decisions including scheduling optimization, cost analysis, and procurement.
An AI-powered customer acquisition engine that helps enterprises precisely target customers and auto-generate marketing content, achieving full-chain automation from lead mining to conversion analysis.
A 24/7 intelligent customer service system integrating NLP and sentiment analysis to deliver efficient, accurate, and empathetic customer service.
Built on the EdgeOnto four-layer architecture, embedding AI Agents into the daily operations of the manufacturing shop floor.
The shop floor supervisor notices Line 3 OEE drop from 85% to 62% — cross-referencing MES, ERP, IoT data and calling the shift lead across 4 systems takes 2-4 hours on average, and the root cause may still be missed.
The shop floor supervisor asks in natural language: 「Why did Line 3 OEE drop today?」Agent Engine understands the intent, pulls MES downtime records, ERP material data, IoT device parameters, and WMS mold change records via MCP Hub; the OEE Analysis Skill calculates loss factors; Knowledge Engine retrieves similar historical cases — within 10 seconds, a complete root cause report is output: 「The OEE drop was primarily caused by a 14:20 mold change exceeding standard time (45min vs 25min standard). Related factor: Material Batch A had higher viscosity causing demolding difficulty.」
Root cause time slashed from 2-4 hours to 10 seconds. Cross-system correlation at 100%. Response moves from hours to seconds. OEE closed-loop time compressed 90%+.
After creating a CRM order, a planner manually enters it into ERP, a scheduler plans it in MES, and warehouse staff reserve materials in WMS — spanning 4 systems and 3+ roles, taking 4-6 hours with error-prone handoffs.
After the salesperson confirms the order in CRM, Agent Engine automatically triggers the cross-system workflow: MCP Hub reads order details from CRM → writes to ERP to create a production order → queues it in MES → notifies WMS to lock material inventory. The entire process runs in Assisted Create mode — Agent generates operation previews for each step, with human confirmation required at critical decision points (e.g., delivery date conflicts). Order processing shifts from 「human-driven」 to 「Agent-driven」.
Order processing reduced from 4-6 hours to 3 minutes (with human confirmation). Error rate drops to zero. Cross-system consistency at 100%. Staff freed for high-value work.
New employee training takes 2+ weeks of one-on-one mentoring, consuming skilled worker time. Process documents are scattered — new hires struggle to find information, leading to high error rates.
New employees ask the Knowledge Engine in natural language on day one: 「What are the standard steps for SMT feeder changeover?」 「What is the quality tolerance range for this process?」 Agent retrieves the latest SOPs, operation videos, and historical knowledge base in real time for precise answers. The AI Digital Human provides voice-guided step-by-step assistance for first-time operations, with the system comparing operational data against standards in real time and correcting deviations immediately.
Onboarding reduced from 2 weeks to 3 days. First-month error rate drops 60%. Mentor time freed by 80%. Training costs reduced 50%+.
Quick deployment with low startup cost, ideal for small and medium-sized manufacturers. Maintenance-free, ready to use out of the box, with continuous feature updates and security hardening. Encrypted data transmission with tenant space isolation.
Recommended: Small & medium factoriesBalancing performance and security — core data (production parameters, process formulas) stays in the local private environment, while non-sensitive services run in the cloud. Enterprise-grade API gateway for unified traffic management, meeting tiered data governance requirements.
Recommended: Enterprises with tiered data governanceAll services deployed on internal servers or private cloud — data never leaves the factory, meeting the strictest compliance and security requirements. Supports deep integration with existing IT/OT systems, with custom model fine-tuning and tailored development.
Recommended: Large enterprises / regulated industries