
Security News
Deno Goes the Way of the Dinosaur
Cloudflare is acquiring Deno’s team, which will provide another year of bug fixes and security updates before ending support for the runtime.
dwelldelta-mcp
Advanced tools
Is an Australian rent fair? A rent check on the ABS rents index, rents by area for four states, NSW median sale prices, and sourced property and lending indicators, for AI agents.
An MCP server that lets an AI assistant check a weekly rent against the ABS rents index, look up median rents by area in four states and NSW median sale prices, and quote sourced Australian property and lending indicators. From DwellDelta, a service of Xcelerent Pty Ltd, built on public data.
Six read-only tools, and one that adds the user's own rent when they ask:
| Tool | What it answers |
|---|---|
rent_check | a weekly rent, and any proposed new rent, against the ABS rents index for the capital city since the lease began (the same arithmetic as the rent check at tools.dwelldelta.com.au) |
market_indicators | the latest RBA cash rate, unemployment, wage growth, rents for all eight capitals, NSW lending, lending standards and arrears, approvals and completions - each with period, source, licence and limits; filter by geography |
vic_rents | Victorian rents by suburb group from Homes Victoria's Rental Report: the median weekly rent of new leases, the lease count and the change on a year before, for 146 suburb groups and 13 regions; filter by suburb_group or region (from 0.5.0) |
area_rents | median weekly rent of new bonds, all dwellings together, with the bond count and the change on a year before, for one state per call: NSW by postcode or council (NSW Department of Communities and Justice), Queensland by suburb or council (Residential Tenancies Authority), WA by locality (calculated by DwellDelta from the Government of Western Australia's bond records); each CC BY 4.0 (from 0.6.0) |
nsw_sale_prices | NSW median sale prices by postcode or council from the NSW Department of Communities and Justice's Rent and Sales Report (CC BY 4.0): the median, the quartiles, the number of sales and the change on a year before, each for all dwellings, houses (non-strata) and units and town houses (strata). Medians of homes sold - never what any one home sold for; a line a few very large sales have set is marked UNUSUAL (from 0.7.0) |
data_sources | the sources, their licences, and what DwellDelta cannot answer |
contribute_rent | adds the user's own rent, anonymously, to DwellDelta's suburb benchmarks - only when the user asks, and only after they have read and agreed to the consent wording (from 0.5.0) |
It runs on your own machine over stdio. It records nothing and sends nothing about you unless you choose to add
your rent. At start it fetches DwellDelta's published files from tools.dwelldelta.com.au - the week's indicators,
the rents and sale prices it answers from, eight public files in all, the same for everyone; nothing about you or
your questions is sent - and otherwise uses the copies it ships with. Set DWELLDELTA_OFFLINE=1 and none of those
eight files is fetched; adding a rent with contribute_rent still sends it, as you asked. Nothing it says is a
forecast, a valuation or advice.
contribute_rent is the one tool that can send anything, and it takes two calls. The first stores and sends
nothing: it returns the consent wording, the Contributor Terms and
Privacy Notice, and a token. Only when the assistant has shown you
that wording and you have said yes does it call again, with the token. What is then stored, in AWS Sydney, is the
suburb, home type, bedrooms, rent, lease month and lease kind, and whether it is a rent you pay or the rent of a
property you own - never an address, name or email. You get a deletion code; email it to lingtao@xcelerent.com
and the record is deleted. Only your own rent may be added: an assistant must not add a figure it read on a
website or a listing.
nsw_sale_prices gives DCJ's medians for the dwellings sold in a quarter - not asking prices, not valuations,
and never the price or value of any one home; no individual sale is in the data. Every answer names the dwelling
type, because a sale price without it answers nobody's question, and carries the quarter, DCJ's source line and
its limits. DCJ reports a quarter when about 80% of its sales have been notified, so the counts are incomplete
and the sales quarter runs behind the rents quarter. DCJ's own table has rows that a few very large sales, or
many sales at one identical price, have set: the answer marks those lines UNUSUAL and says not to quote them as
a going price. Where the published file notes that a total is mostly the sales of such a row (its
unusual_part field), releases from 0.8.0 say so on that total's line, and say only that its median and its
change may be moved by that row's sales. Prices in the structured result stay in DCJ's unit, thousands of dollars, and unit says so.
The feed is public at https://tools.dwelldelta.com.au/market/nsw-sales.json.
Claude Desktop - add to claude_desktop_config.json:
{
"mcpServers": {
"dwelldelta": {
"command": "npx",
"args": ["-y", "dwelldelta-mcp"]
}
}
}
Claude Code: claude mcp add dwelldelta -- npx -y dwelldelta-mcp. Any other MCP client: run npx -y dwelldelta-mcp
as a stdio server. Node 18 or later.
Then ask: "Is a rise from $700 to $770 a week in Sydney since July 2025 in line with rents?", "What is the median rent of new leases in postcode 2148?" or "What is NSW unemployment doing?"
area_rents answers for one state at a time, because the three publishers measure in their own ways and the WA
medians are DwellDelta's own calculation from the Government's one-record-per-bond data. They must not be added
or ranked as one series. Every answer carries that state's quarter, its source line, what DwellDelta selected or
calculated, and its limits. These are medians of new bonds, all dwellings together - not advertised rents, not
what sitting tenants pay, and not the rent of a home of a given size. Where a publisher reports nothing for a
small area the answer says so; it never shows a zero. The feeds are public at
https://tools.dwelldelta.com.au/market/nsw-rents.json, qld-rents.json and wa-rents.json; the server reads them
at start the way it reads the indicators, and DWELLDELTA_STATE_RENTS_DIR points it at a folder of copies.
Victoria, measured differently again, stays in vic_rents.
Give area_rents a dwelling_type, a number of bedrooms, or both, and for NSW and Queensland it answers with
the publisher's own figure for that kind of home - a two-bedroom unit, a three-bedroom house - under the
publisher's own label. One of the two is enough: a type gives every size the publisher reports for it, a size
every type. A combination with no published median is said to have none; it is never filled from the
all-dwellings figure, a neighbouring area or a wider group. Many of these figures rest on few bonds, and each
small sample is marked. The Residential Tenancies Authority does not define its labels: reading "Flat 2" as a
two-bedroom flat is DwellDelta's reading, and every Queensland answer says so. WA's bond records carry neither
dwelling type nor bedrooms, so WA has no such figures. DCJ's largest size is "4 or more Bedrooms": a number above
four is answered under that label, and the answer says it is not for that number alone. A number followed by a
plus means "or more", which only that one size has; anything else with a plus is refused, not answered as exactly
that many.
The two files (https://tools.dwelldelta.com.au/market/nsw-rents-by-type.json and qld-rents-by-type.json) are
large and are not bundled. The server asks for them once at start, without waiting for them, and never in
response to a question. Offline, put copies in the folder DWELLDELTA_STATE_RENTS_DIR names; without them a
by-type question fails with the next step, and every other answer is unaffected.
It is listed in the MCP registry as au.com.dwelldelta/dwelldelta.
Every call returns text to quote and a structured result: the figures, plus sources (each with its licence and
attribution), data_as_of, snapshot_built, snapshot_origin (published for this week's file from
tools.dwelldelta.com.au, bundled for the copy in the package), server_version, retrieved_at, and stale /
stale_reason.
Every failure is marked isError and ends with Next step: - an unknown city or geography names the valid ones.
The indicators are a weekly snapshot, rebuilt every Monday and published at
https://tools.dwelldelta.com.au/market/indicators.json. The server uses that file when it is newer than the copy it
ships with, and the bundled copy when offline. Once a snapshot is more than 10 days old every answer says so, in
words and in the structured result, rather than passing an old figure off as current. Set DWELLDELTA_INDICATORS
to read another snapshot file.
The code is MIT (see LICENSE). The data is not: each figure keeps its source's licence and attribution, which every answer states - for example ABS data is CC BY 4.0, "Based on Australian Bureau of Statistics data".
npm test here starts the server and calls the six read-only tools, and the first (consent) call of
contribute_rent, over MCP; the repository's root npm test holds the
tool answers themselves and checks that vendor/ matches its sources. npm pack runs scripts/vendor.mjs first.
FAQs
Is an Australian rent fair? A rent check on the ABS rents index, rents by area for four states, NSW median sale prices, and sourced property and lending indicators, for AI agents.
The npm package dwelldelta-mcp receives a total of 0 weekly downloads. As such, dwelldelta-mcp popularity was classified as not popular.
We found that dwelldelta-mcp demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
Cloudflare is acquiring Deno’s team, which will provide another year of bug fixes and security updates before ending support for the runtime.

Security News
Anthropic’s free OSS Scanner uses Mythos and other Claude models to find open source vulnerabilities and send maintainers reports with proposed fixes.

Security News
On Front Lines’ Early Adopters: CISO podcast, Andrew Becherer shares lessons from Datadog, weighs the long-term costs of AI-built tools, and explains how founders can actually gain a CISO’s trust.