This site has an MCP endpoint, so an AI assistant can read the register directly and answer with citations back to the tabled pages. It is read-only and needs no account.
Claude — opens the Add-connector dialog with the name and URL filled in; click Add. Works on claude.ai web and Claude Desktop.
Cursor — opens Cursor and installs the server; confirm the prompt.
ChatGPT — has no one-click form for custom connectors, so this opens Settings → Connectors: enable Developer mode, click Create, set the server URL to https://victoriavotes2026.findandseek.app/mcp.
Claude Code & other MCP clients — copy the command, or add it as a remote MCP server (HTTP transport) in any client with the endpoint below.
Endpoint
https://victoriavotes2026.findandseek.app/mcp
What it can answer
Which members declared gifts from Racing Victoria?
Who declares the most land entries, and how many are held through a trust?
List every complimentary membership declared, grouped by organisation.
What did the member for Croydon declare, with page citations?
Which ministers declared gifts from sporting bodies?
Compare declared company interests by party.
What it returns
Each answer carries the member, the form part, the page number and a link to that page of the tabled PDF, so anything the assistant says can be checked. Seven read-only tools cover search, one member's full return, rankings and donor reach: there are no free-form queries and nothing can be written.
Why there is no bulk download
Every declaration is published in full on this site, and each line links to the page of the tabled document it was printed on — so any figure here can be checked against the original, which is what a spreadsheet of our numbers could never prove on its own. If you need the tables for research or a newsroom, write to corrections@findandseek.app and we will send them.
What we log
We count which tools are called and keep the search terms, so we can see what people come here to find out. We do not log who you are, and the endpoint has no accounts, no keys and no sign-in.
Built with FindandSeek: the engine reads the returns once at ingest and answers from what it has read, rather than re-reading documents for every question.