Skip to main content
MCP Server

MCP Server for
Industrial AI Assistants

Proxus MCP Server connects your chosen AI assistants to unified namespace and telemetry data, enabling grounded answers with strict read-only access and full governance.

Why Proxus MCP Server?

MCP (Model Context Protocol) is the standard way to connect AI assistants to your systems. Proxus MCP Server provides a secure bridge to industrial data with strict scope controls. See how it works.

  • Secure authentication—JWT caching and session management built-in.
  • OData proxy—controlled access to Device, Tag, Rule, and Function entities.
  • Telemetry querying—read-only ClickHouse access with injection protection.
  • Schema discovery—auto-generated documentation for entities and data models.
  • Governance—rate limiting, IP whitelisting, and full audit logging.

What stays protected

  • No direct PLC/SCADA access
  • Read-only access to telemetry DB and UNS topics you approve
  • Outbound-only; no inbound control paths
  • Auditability of queries and actions
Proxus MCP Server is a safe reader of your data, not a writer to your control layer.
Where are the bottlenecks in Line 3 right now?
AI Assistant (via MCP)
Based on live OEE data, Press Machine #04 has stopped 3 times in the last hour due to "Material Jam". This has reduced Line 3 efficiency by 12%.
Generate a maintenance ticket for it.
AI Assistant (via MCP)
Done. Ticket #MT-4029 created in SAP and assigned to the on-duty technician.

How it works

Assistants connect through MCP. Proxus provides read-only tools for telemetry, UNS, and entity metadata. You decide the scope. Read the MCP Server docs for setup steps.

  1. Select which telemetry tables, UNS topics, and entities are visible.
  2. Assistants authenticate via MCP; no direct DB credentials are shared.
  3. Queries run read-only against telemetry data; results are contextualized with UNS.
  4. Optional actions (tickets/notifications) go through approved endpoints.
  5. Audit logs capture who asked what and when.

Allowed data paths

  • Telemetry database (read-only, scoped tables/time ranges)
  • UNS topics (read-only)
  • Metadata: sites, lines, assets for better grounding
No control writes. No PLC/SCADA sessions. Assistants stay on the data side only.

Use Cases

Give teams instant answers while keeping production safe and read-only. Review FAQ for governance details.

Operations Q&A

Ask for current throughput, quality yields, or downtime reasons per line or shift.

Maintenance triage

Summarize recent faults and suggest the right work order in ERP/CMMS.

Energy & cost insight

Answer consumption and cost questions from telemetry without custom reports.

Shift handover

Generate concise summaries of alarms, stoppages, and KPIs for the next crew.

Security and governance

AI access is explicitly limited. You choose the data and time ranges; everything is logged.

  • Scoped tools—only the telemetry/UNS readers you enable are visible to assistants.
  • No credentials leak—assistants never receive DB passwords; access is proxied.
  • Audit trails—questions and actions are recorded for review.

What AI can’t do

  • No control writes or setpoint changes
  • No direct PLC/SCADA connectivity
  • No access outside approved telemetry/UNS scope
Safe by default; assistants remain consumers of approved data only.

FAQ

Common questions on safety, data access, and governance.

Does AI get direct DB credentials?

No. Access is proxied through read-only tools; credentials are never shared.

Can AI write to control systems?

No. It only reads approved telemetry/UNS data. Optional actions are limited to governed endpoints like ticketing.

How do we limit scope?

Choose which topics, tables, and time windows are exposed. Everything else stays invisible.

Is access audited?

Yes. Queries and actions are logged for review.

Ready to let your team talk to the factory safely?

Turn on read-only AI access to telemetry and UNS. Keep control logic isolated, keep answers fast.

Read MCP Server Docs Talk to an Engineer