What Is MCP and Why Does It Matter for Construction Software?
MCP is the open standard that lets ChatGPT, Claude, Microsoft Copilot and other AI assistants work with your construction software's live records. What it is in plain English, what it can and can't do for a trade contractor, and what to ask your vendor.
What Is MCP and Why Does It Matter for Construction Software?#
MCP (Model Context Protocol) is an open standard that lets an AI assistant such as ChatGPT, Claude, Microsoft Copilot or Cursor plug into other software and work with its live records, not just the text you paste in. For a trade contractor, it is the difference between an AI that can tell you how the hospital job is tracking against budget from your own timesheets and cost codes, and one that can only guess. Before you count on it, check three things: whether your construction software has its own MCP server, whether that server signs in as you and respects your permissions, and whether your AI plan lets you add outside connectors.
What is MCP, in plain English?#
The official MCP site puts it this way: "Think of MCP like a USB-C port for AI applications." One plug shape, and any assistant that supports it can connect to any software that offers it.
Before MCP, every AI tool needed its own custom link to every piece of software. With MCP, your software vendor builds one connector, called an MCP server, and it works with every assistant that speaks the standard.
Anthropic, the company behind Claude, released MCP as open source in November 2024. In December 2025 it handed MCP to the Agentic AI Foundation under the Linux Foundation, co-founded by Anthropic, Block and OpenAI, so no single AI company owns it. ChatGPT, Microsoft Copilot, Gemini and Cursor had all adopted it by then.
Put simply: a chatbot answers. An agent acts. MCP is how an assistant gets from the first to the second, because it gives the AI tools it can use in your software, like "look up this job's cost report" or "create an estimate."
Why does it matter for a trade contractor?#
Because the questions you actually ask live in your own records. Take a normal Monday. You want to know which jobs burned more labour last week than they earned, who is scheduled tomorrow with a ticket that expires this month, and whether the GC paid the October progress claim. A general chatbot can't answer any of that. It has never seen your timesheets, your schedule or your invoices. So someone exports a report, pastes it into the chat, and hopes the export was current.
With an MCP connection to your construction software, you ask in plain English and the assistant looks it up. The answer comes from the same records your office works from.
Most contractors haven't started. Statistics Canada found that 19.2% of Canadian businesses used AI to produce goods or deliver services in the 12 months before its second-quarter 2026 survey. In construction it was 9.2%, among the lowest of any industry. That gap is the opening.
What can an AI do through an MCP connection?#
It depends entirely on what the software vendor offers. An MCP server is a list of tools, and each tool is either:
- Read-only: look things up. "Show me Friday's timesheets for the Lakeshore job." "What's the total outstanding on unpaid invoices?"
- Data-changing: create or update records. "Create an estimate for the clinic from this takeoff." "Set up the cost codes for the new school job from our standard template."
Data-changing tools are where the assistant does the typing for you, and where you want permissions, confirmations and a log.
Which AI assistants can connect to an MCP server?#
Most of the big ones, but where and how varies by assistant and by plan. Checked on the vendors' own pages, October 14, 2026:
| Assistant | Can you add an outside MCP server? | Where it's set up | Plan notes (as published) |
|---|---|---|---|
| Claude (Anthropic) | Yes, as a "custom connector" | Pro and Max: Customize > Connectors. Team and Enterprise: an owner adds it under Organization settings > Connectors, then each member connects under Customize > Connectors | Free (one custom connector), Pro, Max, Team and Enterprise |
| ChatGPT (OpenAI) | Yes, as a custom MCP app or plugin | Admins or owners publish it to the workspace (on Enterprise and Edu, other authorized members can build it first) | Full MCP, including actions that change records, on Business, Enterprise and Edu; Pro accounts can connect with read and fetch only |
| Microsoft Copilot | Yes, through Copilot Studio | Added as a tool on a Copilot Studio agent | MCP in Copilot Studio generally available since May 29, 2025 |
| Cursor | Yes | Settings, or an mcp.json file |
A code editor, so it suits whoever builds your automations |
Plan rules change often. Check your own plan before you buy anything.
Is it safe to connect AI to your job data?#
It can be, if the connector is built properly and you set it up with care. Here is what to look for.
It should sign in as you. A well-built server makes you log in on the software's own sign-in page, and the AI gets a token, not your password. It should then see and change only what your login is allowed to. If your foreman can't see job costs in the app, his assistant shouldn't either.
Prefer the vendor's own server. Community-built MCP servers exist for many popular construction products. Some are good, but they hold your login and touch your project data, and the vendor doesn't stand behind them. Autodesk said as much when it announced its official Revit MCP server in June 2026: with community servers "popping up everywhere," the question is "who do I trust?" Anthropic's help centre warns that custom connectors reach services it hasn't verified, and that "malicious MCP servers may include hidden instructions." Only connect servers from companies you trust.
There should be a record. Ask whether every action the AI takes is logged, with who, what and when.
Know what the AI company does with your data. OpenAI says it doesn't train on ChatGPT Business or Enterprise data by default; on personal plans you can turn off "Improve the model for everyone." Anthropic says it doesn't train on Claude for Work data by default, and that on its consumer plans, raw content from connectors, including MCP servers, isn't used for training unless it's copied into the chat.
A few house rules for your team:
- AI moves first. You make the call. It drafts the estimate, the reply or the report. A person checks it before it goes to a customer, the bank or payroll.
- Each person uses their own login, never a shared account, so permissions and the log mean something.
- Start read-only. Turn on data-changing tools once you trust the answers.
- Know how to disconnect. Removing the connector, or deactivating the user in your software, should cut it off.
What should you ask your construction software vendor?#
- "Do you have your own MCP server, or would we be relying on a community one?"
- "Which AI assistants have you tested it with?"
- "Does it sign in as each user and follow our roles and permissions?"
- "Which tools only read, and which ones change records? Can we start read-only?"
- "Is every action logged, and can we see the log?"
- "Can it load data in, like an estimate or a cost-code list, or only answer questions?"
- "What can't it do?"
A vendor with a real connector can show you all seven live in the demo.
Where is MCP not the answer?#
- When the data isn't in a system. If your job costs live in a spreadsheet on the PM's laptop and your timesheets are on paper, there is nothing for the AI to connect to. Fix that first.
- When you need a nightly sync between two systems. Moving invoices and bills between field software and QuickBooks every day is a job for a proper integration, like Appello's two-way QuickBooks Online sync. MCP is for people asking questions and giving instructions.
- When the work needs a trade eye. The takeoff, the price on a risky scope and the call on a change order stay with your people. Takeoff is still human. Everything around it doesn't have to be.
- When nobody will check the output. An assistant can misread a question. If no one reviews figures before they leave the office, keep it to questions until someone will.
Where does Appello fit?#
Appello is construction software for ICI trade contractors with crews in the field, and its MCP server is built into the product as part of Appello Intelligence. If you are a residential builder, a GC that needs whole-building document control, or a one-person shop, the questions above still apply but Appello is not the product for you.
- Appello has validated its connector with ChatGPT, Claude, Microsoft Copilot, Perplexity, Cursor and Grok, and it works with any MCP-enabled AI assistant.
- It has more than 700 tools covering jobs, timesheets and check-ins, schedules, safety forms, certifications, estimates, change orders, invoices, equipment, materials and job cost, including earned vs burned and the WIP report.
- It signs in on Appello's own login page as the user, and every call goes through the same role permissions as the app. You approve read, write or admin access. Every tool call is logged.
- Estimates and cost codes can be imported through it. Paste a takeoff summary or a cost-code list into your assistant and have it build the itemized estimate or set up the job's cost codes. Appello can also open a REST API to customers who want to connect their own systems.
Appello switches the connector on for your company and gives you its address; your QuickBooks Online connection is set up in the Appello web app. See the Appello MCP server page for setup details, or ask to see it in a demo.
FAQ#
Is MCP only for Claude?
No. Anthropic created it, but it is now an open standard under the Linux Foundation, and ChatGPT, Microsoft Copilot, Cursor, Gemini and others have adopted it.
Is MCP the same as an API?
Close, but no. An API is how two programs talk to each other. An MCP server packages a program's functions as tools an AI assistant can find and use on its own. A vendor can offer both.
Can the AI change my data?
Only if the server offers data-changing tools and your login is allowed to make that change. Good connectors label which tools change records, and assistants like ChatGPT ask you to confirm before some actions.
Does my construction software support MCP?
Ask the vendor for their own MCP server and which assistants they've tested. If the only option is a community-built server from a directory, treat it with the same care as giving someone your password.
Sources#
All accessed October 14, 2026.
- Model Context Protocol, What is the Model Context Protocol?
- Anthropic, Introducing the Model Context Protocol (Nov 25, 2024) and Donating MCP and establishing the Agentic AI Foundation (Dec 9, 2025)
- Anthropic, Get started with custom connectors using remote MCP, Is my data used for model training? (consumer), Is my data used for model training? (commercial)
- OpenAI, Creating custom MCP apps in ChatGPT, Business data privacy, Data controls
- Microsoft, MCP is now generally available in Microsoft Copilot Studio (May 29, 2025)
- Cursor, Model Context Protocol
- Autodesk, Introducing the Revit Public MCP Server (Jun 17, 2026)
- Statistics Canada, AI use by businesses in Canada, second quarter of 2026
- Appello Docs, Appello MCP Server
Related reading#
- How to Use ChatGPT to Run a Construction Business
- How to Use Claude in Construction
- Integrating Field Operations with Back-Office Systems
- Data Security Considerations for Construction Software
- What Construction Software Works with QuickBooks for Subcontractors?
- Questions to Ask Before Switching Construction Software
- Appello MCP server
Related Articles
Microsoft Copilot for Trade Contractors: What Can It Do with Your Field Data?
If you already run Microsoft 365, Copilot can sort GC emails, recap coordination meetings and pick apart a spreadsheet. It can't see your timesheets, cost codes or safety forms until you connect your field software over MCP. Here's which Copilot you have, what it does out of the box, and the two ways to connect it to live job data.
How to Use Claude in Construction: Estimates, Cost Codes and Timesheets by Chat
How a trade contractor sets up Claude: a Project with your company context, connectors to email and files, a skill for your bid-intake routine, a schedule, and a connector to your job data so estimates, cost codes and timesheets can be handled by chat.
How to Use ChatGPT to Run a Construction Business (and Connect It to Your Job Data)
A practical guide for trade contractors: what to hand ChatGPT first, how to give it context, connect your email, files and job data, put it on a schedule, and which plan you need. Plus what it should never do on its own.
Ready to Transform Your Operations?
See how Appello can help your ICI contracting business.
Book a Free Demo