Can it change data or only read it?

The connection is built around operational visibility. Anything it can do is bounded by the tools exposed and the permissions of the connection you authorised, and the connection can be revoked from Settings at any time.

Is our data used to train a model?

The connection queries your data in response to your questions. What the AI provider does with conversation content is governed by your agreement with them — Anthropic and OpenAI both offer business terms that exclude training. Rostery neither sends data unprompted nor stores your conversations.

Connect Claude & ChatGPT

Every operational question you ask your system is a report someone has to build or a screen someone has to know about. Connecting an AI assistant to Rostery lets you ask in plain English instead — who is unrostered tomorrow, which participants have no notes this week, what did we deliver in June.

Connect claude chatgpt — how it works

Rostery exposes a per-organisation connection using the Model Context Protocol, secured by OAuth. Connect Claude or ChatGPT from Settings, authorise it against your organisation, and the assistant can query your live data within the permissions you granted.

The connection is deliberately operational only. Clients, rosters, progress notes and incidents are reachable; money is not. Timesheets, invoices and payments are outside the tool set entirely, because an assistant that can read financial records is a different risk conversation.

Why it matters for NDIS providers

The questions that go unasked are the ones where building a report costs more than the answer is worth. Lowering that cost changes which questions get asked, which is where most of the value is.

Plain English, live data: ask the question rather than building a report to answer it.
Operational only, on purpose: money, payroll and invoicing are outside the connection.
Scoped to your organisation: authorised per organisation with your existing permissions.

What's included

Per-organisation connection. Your own endpoint, authorised by OAuth from Settings.
Claude and ChatGPT support. Standard Model Context Protocol, so both connect the same way.
Operational tool set. Clients, rosters, progress notes and incidents.
Financial data excluded. Timesheets, invoices and payments are not exposed.
Revocable. Disconnect from Settings and the access ends.

Who it's for

Service managers and coordinators who ask operational questions daily, and anyone who currently waits on someone else to run a report. It suits organisations already using Claude or ChatGPT, since the connection extends a tool they have rather than introducing a new one.

Getting started

Connect Claude & ChatGPT is included in your Rostery plan and works the moment your data is in — there is no separate module to buy or set up. Book a demo and we will show it on your own workflows, and use Rostery's “Smart Switch” data migration to bring your existing clients, staff, shifts and notes across from your current software in minutes.

Terms used on this page

Related capabilities

Source

The rules this feature works to are set by NDIS Quality and Safeguards Commission, which is the authority on them and is updated more often than any page here.

Why teams love Connect Claude & ChatGPT

Per-organisation and consented

Each org connects its own account through OAuth — nothing is shared across tenants.

Operational data only

Clients, roster, notes and incidents. Money, payroll, invoices and settings are deliberately out of reach.

Nothing to install

A remote connection, so it works with Claude and ChatGPT as they are.

Questions instead of reports

The cost of asking drops, so more gets asked.

A deliberately narrow blast radius

Operational data only, revocable from Settings.

Connect Claude & ChatGPT — questions

No. The exposed tools cover clients, rosters, progress notes and incidents. Timesheets, invoices, payments and payroll are outside the tool set entirely — not permission-gated but absent, which is a stronger guarantee.