Ask your assistant.
Not your reports.
AssetOS is building an MCP server, so the AI assistant your team already uses can read your asset register, work orders and maintenance schedules and answer in plain language. No portal, no report builder, no export to a spreadsheet. In build now.
Quick facts
Your CMMS holds the answer.
The reporting screen decides how long it takes to get it.
The reporting screen was designed for a monthly review, and most questions are not monthly. Anything outside the shape of an existing report means filters, an export, and starting over when the grouping turns out wrong.
- ✕The person who knows how to build the report is the person who built it, and they are on leave.
- ✕Half the questions a supervisor has never get asked, because the answer costs more than it is worth.
- ✕An auditor wants a number now, and the reporting screen was designed for a monthly review.
- ✕Every follow-up question means starting the export again from the top.
It answers questions and does nothing else.
Your assistant connects to AssetOS, reads the records your account is allowed to read, and answers in the words you asked the question in.
Questions, not report builders
“Which pumps at Site 3 missed their last PM?” gets an answer, then a follow-up gets another. One conversation covers what used to take several trips through the reporting screen.
Your role is the boundary
Each person authenticates as themselves over OAuth. The server inherits that user's AssetOS role, so it reaches exactly what they can reach and nothing beyond it.
Read-only, deliberately
Version one reads. Write tools will come with their own scopes and their own consent step, because an assistant closing the wrong job at 2am is worse than no integration at all.
Five steps become one sentence.
Your data stays where it is. The distance between having a question and having the answer gets shorter, and more questions get asked once asking is cheap.
- Follow-up questions cost nothing, so they get asked
- No reporting screen to learn, so more of the team can ask
- Answers come from your live records, not last month's export
- The spreadsheet export stays available for the people who want it
- 01Log in to the portal
- 02Find the right report
- 03Set the filters and dates
- 04Export to a spreadsheet
- 05Repeat for the next question
It signs in as you, and only as you.
There is no service account and no shared key that quietly reads the whole tenant. Each person connects with their own AssetOS login, and the server carries that person's permissions into every question they ask.
- Per-user OAuth — the same consent pattern as Slack and Teams
- Role inherited from AssetOS, not re-declared somewhere new
- A site your account cannot see stays invisible to your assistant
- Disconnect from AssetOS and access stops immediately
When you ask a question, the records the server returns go to the AI assistant you connected. That is how it answers. Any vendor telling you their AI integration never shows your data to the AI is describing something else.
- You choose the assistant. AssetOS does not pick one for you.
- AssetOS does not call a model on your behalf — the server returns your records and stops.
- AssetOS does not train any model on your data.
- Not acceptable to your organisation? Use the REST API instead. It is live today.
What it won't do — said now, not after you buy.
This is the boundary of version one, in plain terms. If something on this list is the reason you'd switch, tell us — it moves up the queue.
Create or close work orders
Version one reads and nothing more. Write tools will come with their own scopes and their own consent step, because an assistant closing a job nobody did is worse than no integration at all.
Log every question to the audit trail
Not yet. The AssetOS audit trail covers work orders and asset changes as it always has; per-question logging for MCP reads is on the list and is not in version one. We would rather write that here than let you assume it.
Make the assistant's answer our answer
The server returns your records accurately. How an assistant summarises them is the assistant's behaviour, not ours. For anything you are signing your name to, open the record.
Work without an AI assistant
This integration is only useful if your team already uses one. If they don't, nothing on this page is a reason to start — the reporting screens and the API are not going anywhere.
The data is already reachable.
The MCP server changes who can ask and how fast. It is not what makes the records available — they are available now.
REST API
The same endpoints our mobile app uses. Full read access to assets, work orders and schedules, and plenty of teams already wire it into their own dashboards.
Webhooks
Push events out as they happen rather than polling for them. Work order status changes, asset changes, and the rest.
Slack and Teams
Live today. Alerts land in the channel your team already watches, with acknowledge, comment and close from the message itself.
The first questions people ask
will decide what we build.
We have a list of the questions we think maintenance teams want answered. It is a guess until real people ask real questions against real data. Early access decides which tools version one ships with — and which ones we were wrong about.
MCP and AssetOS — the questions we get asked.
Does AssetOS have an MCP server?
Not yet — the AssetOS MCP server is in build. It will let an AI assistant read your asset register, work orders, preventive maintenance schedules and service history, and answer questions about them in plain language. Early access customers get it first. Today the same data is reachable through the AssetOS REST API and webhooks.
What is MCP?
The Model Context Protocol is an open standard for connecting AI assistants to external systems. An MCP server exposes a defined set of tools the assistant is allowed to call. It is not a chatbot bolted onto a product — it is a connection your assistant makes to real data, so the answers come from your records rather than from a model's guess.
Can the MCP server change anything in AssetOS?
No. Version one is read-only. It answers questions and it does not create, close or delete anything. Write tools will come later, with their own scopes and their own consent step.
Can it see data my role cannot see?
No. Each user authenticates over OAuth and the server inherits that user's AssetOS role. If your account cannot see a site, neither can your assistant. Disconnect from AssetOS and access stops immediately.
Where does my data go when I ask a question?
To the AI assistant you chose to connect, and nowhere else. That is how the assistant answers — it reads the records the server returns. AssetOS does not call a model on your behalf and does not train any model on your data. If sending asset data to an assistant provider is not acceptable to your organisation, this integration is not for you, and the REST API is.
Which AI assistants will it work with?
MCP is an open standard, so any MCP-compatible client can connect. We will name the clients we have actually tested against when it ships, rather than listing every assistant that theoretically works.
Does this replace the AssetOS API?
No. The REST API and webhooks are for systems you build against. The MCP server is for questions people ask. Both stay, and the API is available today.