Model Context Protocol

Up-to-date lender policieswhere you work.

Connect CreditPolicy to Claude, ChatGPT, Cursor or any MCP client, and it searches your Australian lender policy library mid-conversation — cited to the source document, grounded in your own files. No tab switching. No guessing.

https://creditpolicy.ai/api/mcp
Works with any MCP client Cited to source Revoke anytime
Claude
CreditPolicy connected
Message Claude

One server, every assistant

Connect it once. Use it wherever you work.

MCP is an open standard, so CreditPolicy isn't tied to one assistant. The same endpoint works across every major client.

creditpolicy.ai/api/mcp

Claudeclaude.ai, Desktop & CodeOne-click sign-in
ChatGPTApps & ConnectorsOne-click sign-in
GrokRemote MCP serverAccess key
PerplexityRemote MCP serverAccess key
CursorRemote MCP serverAccess key
VS CodeCopilot agent modeAccess key
Gemini CLINative MCP supportAccess key
WindsurfRemote MCP serverAccess key

Anything that speaks MCP works — including automation platforms like Make, Zapier and n8n, and agents you build yourself.

0

Lender documents

0

Policy libraries

0ms

Typical search

0

Tools exposed

Why it matters

The answer arrives where the work is.

CreditPolicy already answers policy questions. Connecting it over MCP means those answers turn up inside whatever you're already doing.

Policy answers where you already work

Stop leaving the conversation to check policy.

You're already in Claude or ChatGPT drafting the client email. Ask the policy question in the same breath — max LVR, casual income, HECS treatment — and keep writing.

It reads your library, not the internet

It can't invent a lender policy.

Every answer comes back with the source document and a relevance score, pulled from your own policy library. If the library has nothing, it says so instead of guessing.

Chain it with everything else

One prompt, whole task.

"Check whether Macquarie takes 100% of bonus income, then draft the file note." It checks the policy and writes the note — you're not the middleman copying between tabs.

Pre-vet before you lodge

Run the panel without opening the panel.

Describe a deal in plain English and get lender-by-lender assessments with the criteria met and unmet, so dead-end submissions die in the conversation, not at credit.

What it can do

Your whole workflow,one connection.

This isn't a search box bolted onto a chatbot. Twelve tools span the policy library, your client book and your deal pipeline — and your assistant chains them together in a single conversation: check the policy, vet the deal, update the client, open the deal, draft the email.

Ask the policy library anything

Your assistant searches the lender documents themselves and reasons over the exact wording — not a summary, not the internet.

  • Any policy question answered from the source document, cited by name
  • Auto-routed across six libraries — residential, commercial, asset, private, personal and compliance
  • Two to four lenders side by side on the same question, each in their own words
  • If the library has nothing, it says “not in library” — it never fills the gap with a guess

Pre-vet deals before you lodge

Describe a deal in plain English and the scenario engine runs it across the whole panel while you keep talking.

  • Lender-by-lender verdicts — strong, possible or unlikely
  • Every criterion met and unmet, spelled out per lender
  • Runs are saved to your scenario history in the app automatically
  • Dead-end submissions die in the conversation, not at credit

Run your client book by talking

The CRM is in the conversation too. Paste a fact find and the record builds itself; ask about a client and the file opens.

  • Find any client by name, business, email or ABN and pull their full profile
  • Create or update clients — income, liabilities, property and business details included
  • Open deals, move them through submitted, approved and settled
  • The conversation files itself against the client and deal it was about

Wired in properly

One connection, then it behaves like part of your account — same credits, same history, same guardrails.

  • Sign in with OAuth from claude.ai or ChatGPT — no keys to paste
  • Personal cpk_ keys for Claude Desktop, Cursor and automation that can't do OAuth
  • Every MCP conversation lands in Saved chats beside your web and email ones
  • Policy library is read-only, nothing can ever be deleted, access revocable anytime

And it chains — one ask, several tools.

Create the client, run the scenario against that client, open the deal. You say it once; your assistant works out which tools to call and in what order.

Here's the fact find from my 2pm — set them up and tell me who'd approve.

create_clientassess_scenario

The client record is built from the pasted email — income, debts, property and all — then their numbers run across the panel. Verdicts come back lender by lender, and the run is saved to your history.

Would Liberty take Sarah's deal if she pushed to 90% LVR?

list_clientsget_clientsearch_policy

Sarah's actual financials are pulled from your book and read against Liberty's own policy wording — the answer is about her file, not a hypothetical.

She's going with Resimac — open the deal and draft the follow-up email.

create_deal

The deal is opened and filed against her record with the conversation attached, and the email is drafted in the same reply. You never left the chat.

What your assistant gets

Twelve tools. Every answer cited.

1
search_policy

Search every lender policy in the CreditPolicy library and get the exact wording back, cited.

2
list_lenders

See which lenders are covered before asking.

3
compare_lenders

Put two to four lenders side by side on the same question.

4
assess_scenario

Run a whole deal across the panel and see who's likely to approve.

5
list_clients

Find any of your clients by name, business, email or ABN.

6
get_client

Pull a client's full profile into the conversation and talk it through.

7
create_client

Create a client from what you tell it — including income, debts and property.

8
update_client

Update an existing client — only the fields it's told to change.

9
add_client_financials

Fill in a client's income, liabilities, property and business details later.

10
list_deals

See what's on for a client, or across your whole book.

11
create_deal

Open a deal against a client, so the conversation is filed with it.

12
update_deal

Move a deal along — submitted, approved, settled.

The policy library is read-only. It can create and update your own clients and deals and save scenario runs to your history — it can never delete anything, and it has no access to your forms or billing.

Setup

Connected in about a minute.

1

Get your access key

Checking your account…
2

Add it to your assistant

  1. 1. claude.ai → Settings → Connectors → Add custom connector
  2. 2. Paste the server URL, click Connect.
  3. 3. Sign in to CreditPolicy and approve.
https://creditpolicy.ai/api/mcp

No key needed here — you sign in through CreditPolicy and approve, the same way you'd connect any other app.

Then just ask.

Tap any of these to copy.

Stop tabbing out
to check policy.

Connect once. Ask every policy question from inside the conversation you're already having.