MCP Server
Many people already work with an AI app such as Claude or ChatGPT. The UpBlue MCP Server connects that app to UpBlue, so you can ask about your production data in the same chat where you write reports, analyse Excel files or prepare a meeting:
- “Compare the output of Line 1 and Line 2 last month and make a table for the management meeting.”
- “Which alerts fired this week? Write a short summary for the maintenance team.”
- “Build a dashboard with the uptime of all packing machines.”
MCP (Model Context Protocol) is the open standard that AI apps use to connect to other systems. UpBlue offers one MCP server for all organisations; you log in with your normal UpBlue account.
Server address
https://mcp.upblue.io/mcpThat is the only thing you need. There is no API key or client ID to create: the AI app registers itself and opens the UpBlue login page in your browser.
Step-by-step guides:
- Tutorial: Claude — claude.ai, Claude Desktop and Claude Code
- Tutorial: ChatGPT & OpenAI — ChatGPT and Codex
Supported apps: Claude and ChatGPT, and apps running on your own computer such as Claude Code and Codex. Other web apps are refused at login.
Your own permissions
The AI app works as you, with your own role in each organisation. It sees exactly the organisations you see in UpBlue and can never do more than you can yourself.
| Your role | What the AI app can do |
|---|---|
| Viewer | Read tags, values, events, dashboards, alert rules and event rules. |
| Editor and higher | Also build and change dashboards, connections, rules, collectors, alerts and event rules. |
Want an AI app that can only read? Use it with a user that has the Viewer role.
The AI app is instructed to ask for your approval before every change to connections, rules, collectors, alerts or event rules, and to tell you exactly what will change. Read that question before you say yes: deploying or stopping a connection affects the machine data on site. Most apps also ask for permission themselves before using a tool that writes; leave that switched on.
Dashboard changes are always safe to try: every change creates a new version, and a new dashboard stays a draft until it is published. An earlier version can be restored at any time.
What the AI app can do
| Area | Read | Change |
|---|---|---|
| Organisations | List your organisations, a snapshot of all tags with their latest value | |
| Tags & data | Tags and labels, latest values, history aggregated per minute, hour, day… | Delete a tag with its data |
| Events | Events with reasons and calculations, reason trees | |
| Dashboards | Dashboards, folders, full configuration, version history | Create, update, publish, restore a version |
| Connections | Connections and their state, test a connection, browse OPC UA nodes, cloud connection log | Create, update, delete, deploy, stop |
| Rules | Rules, test a rule (sample values) | Create, update, delete |
| Collectors | Collectors, live status and log | Rename |
| Alerts | Alert rules, test a rule, alert log, alert outputs | Create, update and delete rules and outputs, send a test message |
| Event rules | Event rules | Create, update, delete |
Ask for summaries over a period (“average per hour”, “total per day”) rather than raw values. UpBlue aggregates the data for the AI app, which keeps answers fast and accurate.