WarrantyHub's MCP Server lets the AI assistant your team already uses work directly with your warranty data. Look up a contract, file a claim, import a dealer's spreadsheet, run a report — in plain English, inside your own permissions, with every change previewed before anything is written.
Approved by a company administrator · Scoped to your company and program · Revoke any time
MCP stands for Model Context Protocol — a common standard that lets an AI assistant connect directly and securely to a business system, the same way you might connect a spreadsheet app to your bank account. WarrantyHub's MCP Server is the doorway that lets an assistant you already use work with your WarrantyHub information.
So instead of clicking through screens to answer "is this contract still active, and what's been claimed against it," you just ask. And when there's work to do — a claim to file, forty contracts to import, a report to schedule — you ask for that too.
No query builder, no report designer, no exports to reconcile. These are real requests a connected assistant handles.
A connected assistant can read your warranty operation and act on it — within the permissions you approved, and never without showing you first.
Read
Pull up an individual contract or claim, check coverage and status, and get a summary of everything that's happened on a claim so far — without hunting through screens.
Write · previews first
File a new claim against a contract, move a claim to a new status, update its claim jobs, and add internal notes explaining the decision — with the change shown to you before it lands.
Write · row-by-row
Hand over a dealer's spreadsheet and have it imported. Bad rows fail on their own without taking down the batch, and you get told exactly which ones didn't go through and why.
Read & write
Search your technician network, add new ones one at a time or in bulk, and keep their details current — the same row-by-row safety applies to technician imports.
Write · previews first
Have your plan groups and pricing explained back to you, or set up pricing for a new plan from a SKU list and see the coverage each policy ends up with before you commit to it.
Read & write
Ask a reporting question in plain English, save the answer as a report definition, and put it on a schedule. Large pulls export to CSV in the background with a download link.
Write · previews first
Build the rules that fire on claim events — email the homeowner when a claim is marked Complete, move work along when a status changes — and have the assistant assemble the rule for you.
Write · previews first
Add or adjust the statuses a claim moves through, so the path a claim takes in the system matches the way your team actually works it.
Read & write
Ask which automated emails go out when a claim is approved, then save or edit the template a workflow rule sends — without leaving the conversation.
Write · destructive
Cancel a contract and every coverage period on it. It can’t be undone through the assistant, so it always previews and asks for an explicit confirmation before it proceeds.
Read & write
Have your company and program setup explained back to you — including which one the connection is scoped to and what it’s allowed to do — or change it.
Every capability above is gated by what your administrator approved on the consent screen — a connection only gets what was granted, and nothing more.
The reason to be careful about connecting an AI assistant to your warranty system is exactly the reason we built it this way.
Tools that change data run in preview mode by default. The assistant validates your request, reports exactly what would change — new records, old value to new value, and any rows that would fail — and writes nothing. It only applies the change after you confirm.
On imports of contracts or service technicians, one bad row fails only that row — the rest still process. Re-sending a row that's already correct does nothing, so a retry can't create duplicates.
Cancelling a contract cancels every coverage period on it and can't be undone through the assistant. So it always previews, and always asks for an explicit confirmation before it proceeds.
A connection can only reach the company — and the single program — it was approved for. It acts as the person who approved it, with that person's permissions. It can never see or do more than your own account can.
Every connection has to be approved by a company administrator on a consent screen, which names the application, flags whether it's verified, and lists in plain terms exactly what the connection will be able to do.
Revoke a single agent and its very next request fails. Disable a registered client to cut off every session it holds at once. Or rotate your connect URL to disconnect every agent in every program — effective immediately.
Settings → MCP Access holds your Server URL and copy-and-paste connect instructions for each client. You never type your WarrantyHub password into the AI assistant — you hand it your Server URL, and it sends you back to WarrantyHub to sign in and approve.
ChatGPT connectors require an eligible ChatGPT plan. Older clients that only support local servers can use the bridge command shown on the same page. Administrators: the full connection guide walks through every client step by step.
Book a demo and we'll show you an AI assistant reading your contracts, filing a claim, and running a report — and exactly what it shows you before it writes anything.
Already on WarrantyHub? The MCP Server is installed by support — read the setup guide.