OnboardMe in Claude, ChatGPT and Cursor: what you can do with our MCP server
Connect your AI assistant to OnboardMe and ask it to find clients, draft proposals, start onboarding, chase debtors and more. Here’s how it works, how to connect, and the use cases teams get the most from.
The OnboardMe team 4 October 2026 5 min read
Most of the admin in a service business isn’t hard. It’s just scattered. Finding which proposals are still waiting on a signature, checking who’s 60 days overdue, setting up next year’s engagements, starting onboarding for a new group of entities. Each one is a few minutes of clicking, and they add up.
OnboardMe’s MCP server lets you hand that work to the AI assistant you already use. Connect Claude, ChatGPT, Cursor or any other app that supports the Model Context Protocol, and you can ask for things in plain English:
“Which invoices are more than 60 days overdue, and who’s assigned to chase them?”
The assistant looks it up in your OnboardMe account and answers, or does the work, with the same permissions you have in the app.
What is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to other software. Instead of copying and pasting between OnboardMe and a chat window, the assistant can search your clients, read a proposal or create an eForm itself.
OnboardMe’s MCP server sits on top of our Open API, so it covers the same ground: clients and contacts, leads and deals, proposals, onboarding, identity verification and AML, eForms, bills, payments, statements and debtors.
Use cases by role
Here’s what teams use it for. Every example is a real prompt you can try.
Partners and owners: answers without the reports
- “What’s in our leads pipeline, and what’s the total value of the hot ones?”
- “Which proposals have been sent but not accepted in the last 30 days?”
- “Show me our debtors ageing. Who owes the most in the 90+ bucket?”
- “How much did we collect last month, and how does that compare with the month before?”
Instead of building a report, you ask the question. The assistant pulls the figures from OnboardMe and summarises them.
Client managers: proposals and onboarding in a sentence
- “Create a proposal for Acme Pty Ltd from our bookkeeping template.”
- “Preview renewals for this year’s monthly clients with a 5% increase, then renew the ones I approve.”
- “Start onboarding for the Patel Family Trust with ID checks for both trustees.”
- “Which onboardings are still waiting on documents, and what’s missing?”
The assistant uses your templates, packages and terms, so what it creates looks exactly like something you’d have built yourself.
Compliance: ID and AML on request
- “Send an ID verification to Sam Patel.”
- “Run an Extensive AML check on the directors of Northside Physio.”
- “Download the ID verification certificate for Priya Shah.”
These are paid checks, so the assistant asks you to confirm before it runs one.
Credit control: chase debtors from a chat
- “What’s in my debtor queue today?”
- “Send a reminder to everyone more than 30 days overdue who isn’t on a payment plan.”
- “Add a note to INV-0142: promised to pay on Friday.”
- “Get me a payment link for Harbour Dental’s outstanding invoices.”
It works alongside debtor management. Your workflow still runs on its own; the assistant is there for the exceptions.
Setting up OnboardMe: the jobs nobody wants
This is where an assistant saves the most time.
- “Build an eForm for new trust clients: trustee details, the trust deed upload and a signature.”
- “Turn this Word document into an engagement terms template.”
- “Import our services and prices from this spreadsheet. Show me a dry run first.”
The assistant follows OnboardMe’s own guides for eForm questions and terms formatting, so the results fit the app’s structure. Imports show you what will change before anything is saved.
Everyone: “What do I need to do today?”
- “What’s outstanding for me today?”
- “Summarise my unread notifications and link me to each one.”
It’s a quick way to start the day without opening five screens.
How to connect
Use the address for the region your OnboardMe account is in:
- Australia and New Zealand:
https://anzmcp.onboardme.app/mcp - United Kingdom:
https://ukmcp.onboardme.app/mcp - South Africa:
https://zamcp.onboardme.app/mcp
Claude and ChatGPT. Add OnboardMe as a custom connector using your region’s address. You’ll be taken to OnboardMe to sign in, choose your business, and pick Read-only or Read & write access.
Claude Code. Run:
claude mcp add --transport http onboardme https://anzmcp.onboardme.app/mcp
Cursor and other developer tools. Connect with your OnboardMe API Client ID and secret, sent as request headers. You create these under External API Access in OnboardMe, and the API documentation has the details.
Safe by design
Letting an AI assistant into your client data should make you ask questions. Here’s how OnboardMe keeps it in check:
- Same permissions as the app. The assistant can only do what your user role allows. If you can’t delete a deal in OnboardMe, neither can it.
- Read-only when you want it. Choose Read-only when you connect, and the assistant only ever sees the tools that read data. You can also add
?profile=readonlyto the address to keep a connection read-only. - Confirmation for anything that matters. Sends, bills and paid checks such as ID verification are flagged, so the assistant confirms with you before acting.
- Your region, your data. Each region has its own server, connected to that region’s OnboardMe account.
- Revoke any time. Removing access in OnboardMe cuts the assistant off within about 30 seconds.
What it costs
The MCP server is free, just like the Open API. Anything that’s normally paid in OnboardMe stays paid. An ID verification is charged per check, and AML screening uses requests from your plan, whether you start it in the app or from a chat. Your AI assistant’s own subscription is separate.
Tips for better results
- Name things the way OnboardMe does. “Proposal”, “onboarding”, “eForm” and your template names help the assistant pick the right tool.
- Ask for a preview first. For bulk work like renewals or imports, ask to see what will change before it saves.
- Start read-only. Get comfortable asking questions, then switch to Read & write when you want it to do the work.
See everything in the Open API, or start a 30-day free trial and connect your assistant to your own account.