MCP & connectors
MCP connectors that plug your AI into the tools you already run.
We build connectors that let Claude and other AI assistants work with your business systems: reading live data, answering questions and drafting the next step, safely and with your permissions.
Talk to us about a connectorWhat is MCP?
A plug between your AI and your business.
MCP, the Model Context Protocol, is an open standard for connecting AI assistants to other software. Instead of copying and pasting between systems, your AI can look up a job, check a ticket or pull a report directly, from the source, in the moment you ask.
How it works
How it all stacks up.
Every roadmap builds the same way: from the systems you already run, up to the people who use AI every day.
- 05Govern & maintain →
Your people
Using it every day, confident and supported, with the guardrails in place.
TrainingGuardrailsUpskilling on demand - 04Build & implement →
AI assistants & agents
The tools that do the work: drafting quotes, building reports, answering customers.
QuotingReportingCustomer replies 
Roadmap & strategy →Your business brain
Your knowledge, rules and processes in one place, so every tool works the way you do.
KnowledgeRule bookProcesses- 02Build & implement →
MCP connectors
The secure plugs we build into each system, so your AI reads live data with your permissions.
Connect what is readyBuild what is notRead-only first - 01Roadmap & strategy →
Your systems
What you already run. We start here and work up.
XeroYour ERPMicrosoft 365CRM
How we connect
Connect first. Build only where it is needed.
Every roadmap sorts your systems into four groups. Usually only one of them needs building.
Connect what already has a server
Many tools now ship an official MCP server, such as Xero, HubSpot, Klaviyo, ClickUp and Google Sheets. These are secure sign-ins, not builds, so they can be live quickly.
Build one server for the rest
Systems without a server, often your ERP, inventory or job system, get one server built over all of them, not a connector each. One login model, one set of rules, one view of each customer.
Use the data you already own
If you already have a data warehouse or reporting database, AI reads it where it is. What it usually needs is the missing sources and cost data, not another connector.
Flag what needs a question first
Some systems cannot be connected until someone answers a question: who the vendor is, where files live, what the API allows. We list them openly rather than guess.
Built by us
Connectors we have built and run today.
ERP & business systems
Propella
Customers, stock, orders and pricing from your ERP, so your AI answers from live numbers.
Infusion
Stock, sales, purchasing and jobs, connected to the way your team already works.
Cin7
Inventory, orders and stock levels across every channel you sell through.
WorkflowMax
Jobs, time, quotes and invoices, so your AI can see where every job is up to.
Jobs & field service
AroFlo
Jobs, quotes, clients, sites, the schedule, technicians and invoices. Draft changes only when you switch them on.
Tradie on the Move
Jobs, quotes and invoices, with a daily reporting dashboard built on top.
Print & production
printIQ
Jobs, quotes and production data, connected to the business’s own knowledge.
Customer support
Zendesk
Read-only ticket reporting and analysis, with permissions set per workflow.
Projects & content
Microsoft Planner
Plans, buckets and tasks, both ways. Signs in as you, sees only what you can, and never deletes.
Trello
Boards, lists and cards for status reporting across clients and projects.
Gain
Content approvals and schedules, pulled into one reporting view.
Google Drive
Documents and folders, so your AI works from the files your team already keeps.
Property & mapping
Nearmap
The aerial image for a site address, shown alongside the estimate.
LINZ aerial imagery
Free, government-supplied aerial photos of a site, as a no-cost alternative.
Ready-made
Plus the connectors that already exist, set up properly.
Many tools now offer their own connector. We choose the right ones, set the permissions and connect them to the way your team works.
Four rules
Four rules every connection follows.
Read-only outside the main tool
AI can look things up and draft the next step, but it does not change your records. One rule, enforced in one place.
Events, not polling
Where a tool supports it, connectors react the moment something changes, instead of checking on a timer.
Each tool's own permissions
We use the access controls your tools already have, so people only see what they are allowed to see.
One customer record
Everything points back to your existing source of truth. No second customer database to reconcile later.
The order
The order we build it in.
1
Answer the open questions
Usually one conversation. It costs nothing and stops surprises later.
2
Connect the ready-made servers
Secure sign-ins, scopes checked, limits confirmed. No code.
3
Name a data owner and a usage policy
Every connection signs in as someone. This makes it governable.
4
Build the one server
Starting with your source of truth for customers and orders.
5
Go live
Swap sample data for live data. The screens your team uses do not change.
Looked after
We support it long after it goes live.
Tools change and so does your business. We keep your connectors and AI tools working, update them when your systems or needs change, and add what you need next, with ongoing support for your team.
See govern & maintainUsing a tool that is not listed?
If it has an API, we can usually connect it. Tell us what you run.