The gap nobody was talking about
AI can write code, summarise legal documents, and generate marketing copy. But ask it a simple question about your factory ("What's my MRP status for today?") and it has nothing to work with.
The problem isn't intelligence. It's access. Enterprise systems hold decades of operational data behind proprietary APIs, stored procedures, and vendor-specific schemas. AI agents can't reach it.
We built Pharos to fix that.
What Pharos actually is
Pharos is an MCP (Model Context Protocol) server that sits between your ERP and any AI client. It exposes your manufacturing data as structured tools (schema search, KPI dashboards, MRP analysis, even write operations) that AI agents can call directly.
No middleware. No data pipeline. No ETL jobs running overnight. The AI queries your live production database, and you get answers grounded in real numbers, not hallucinations.
Why MCP, not another REST API?
REST APIs are designed for applications. MCP is designed for AI agents. The difference matters:
- Discovery: An AI agent can ask Pharos "What tools do you have?" and get a structured list. With REST, you need documentation and a developer to read it.
- Context: MCP tools carry descriptions, parameter schemas, and examples that help the AI understand when and how to use them.
- Composability: An agent can chain multiple tools in a single conversation: search the schema, find the right table, query it, then analyse the results.
This isn't a theoretical advantage. At Tru-Trac Rollers, operations staff ask questions in plain English and get answers from their SYSPRO database in seconds. No SQL. No IT ticket. No waiting.
What we learned at Tru-Trac
Tru-Trac Rollers manufactures conveyor components for 80+ countries. They run SYSPRO, Tempo MRP, and a custom analytics warehouse. Before Pharos, getting a cross-system answer required a developer, a SQL query, and patience.
After deploying Pharos:
- Managers query MRP status, customer risk, and production KPIs through natural language
- The AI navigates across all four data sources without being told where to look
- Write operations (labour posting, inventory adjustments) go through confirmation gates: the AI proposes, the human confirms
The deployment took weeks, not months. No changes to the existing ERP. No data migration. Pharos reads the systems as they are.
The vision
Every ERP query should be a conversation. Not a report request. Not a SQL query. Not a three-day turnaround from IT.
Pharos is how we get there, one MCP tool at a time.
Pharos is live in production at Tru-Trac Rollers. If you want to see what it can do with your data, get in touch.