LAKE now speaks MCP, the protocol AI assistants use to reach outside systems. That means the assistant your team already uses can ask the platform directly rather than being fed screenshots and pasted numbers, and can act on what it finds once a person approves.
Why This Matters More Than a Chat Window
Every plant already has an assistant of some kind in use, and every one of them has the same weakness: it only knows what somebody pasted into it. A vibration trend described in a sentence is not a vibration trend, and an assistant reasoning from a summary is reasoning from someone else’s conclusions.
Connecting through MCP removes that step. The assistant asks for the trend, the spectrum or the bearing score and gets the actual figures, from the machine you named, over the window you meant.
What It Can Reach
The interface covers the working set of a reliability engineer, arranged the way an investigation runs. Triage first: which machines need attention, which alarms are active. Then trend and history: telemetry over a window, bearing scores, machine mode history, the events recorded against the machine. Then the detail: a measurement’s signals, its spectrum, peaks near a frequency you name, envelope, cepstrum, waveform.
Around that sit the things that give a number meaning: equipment details, the bearing fitted, recent running speed, process data, the notes and photographs on the machine, and the asset and maintenance records.
Reading and Acting Are Separate
Most of what the interface offers only reads. A small and explicit set changes things, and it is the set you would expect: raise a work order, set or remove an alarm configuration, add a note, update a machine’s health status, create a preventive maintenance schedule.
That separation is deliberate. An assistant that can read everything and change a few well-defined things is a useful colleague. One that can change anything is a liability.
Connecting
An integration connects with an API key. A person connects through OAuth, which is the usual route for a desktop assistant and means the assistant inherits that person’s access rather than holding a shared secret. Either way the session is bound to one facility and can be switched to another, so an account with several plants does not blur them together in one conversation.
Where to Find It
The MCP endpoint is part of LAKE, alongside the webhooks and the scoped API key described on the integrations page. See the developer page for how to connect one, and the Reliability Agent, which is Sensemore’s own client of the same interface.
Frequently asked questions
What is the MCP server for?
It lets an AI assistant reach your Sensemore data directly. MCP is the protocol assistants use to talk to outside systems, so instead of copying numbers out of the platform and pasting them into a chat window, the assistant asks the platform itself and works from what it gets back.
What can an assistant actually do with it?
Read the things a reliability engineer reads: which machines need attention, active alarms, trends, spectra and envelope or cepstrum views of a measurement, bearing scores, machine mode history, equipment details and the machine's event log. It can also act, raising a work order, setting an alarm, adding a note or updating a health status.
Can it change data without me knowing?
The tools that change anything are a small, explicit minority of the set, and they are the ones you would expect: raising a work order, setting an alarm, writing a note, updating a health status. Everything else only reads. What an assistant is allowed to do is bounded by the credentials it was given, exactly like a person.
How does an assistant connect?
Either with an API key for an integration, or through OAuth for a person, which is the usual route for a desktop assistant. The session is tied to one facility at a time and the assistant can switch to another, so a multi-plant account does not get mixed up in a single conversation.
Do I have to use Sensemore's own assistant?
No, and that is the point. The Reliability Agent inside LAKE is one client of this interface; the MCP server is open to whichever assistant your team already works with. Your data stays in the platform and the assistant reaches it with the permissions you granted.