Connect Claude Desktop to HtAG via MCP

Connecting Claude Desktop to HtAG can turn a plain-language property question into a traceable data workflow. The useful part is the hand-off: Claude resolves the requested place, asks HtAG for the matching market segment and presents a dated answer that a person can check.

See the MCP research flow

View the demo screenshot

The interactive concept above shows the connection and query stages. It is not a screenshot of Claude Desktop, and its Richmond figures come from a captured HtAG response. Opening it makes no HtAG API request.

Before you connect

You need Claude Desktop and an HtAG account with MCP access. Sign in to the HtAG Developer Portal, then follow the current HtAG MCP guide for Claude Desktop.

The current guide is the source for the connection URL, authentication method and client-specific steps. Copy its generated or documented values exactly and keep credentials private. Client configuration can change, so avoid reusing a static configuration snippet from an older article.

Verify the connection with a bounded request

Start with one locality and one market segment:

Use the HtAG tools to resolve Richmond VIC 3121.
Then get the market summary for houses, all bedrooms.
Include the reporting period and identify the fields returned by HtAG.

This request has a clear two-step path: locality identity from GET /reference/locality, followed by the selected market record from GET /markets/summary. Richmond VIC 3121 resolves to VIC2167.

Know what a grounded answer looks like

Use the dictionary meanings when Claude explains these fields: Typical Price is HtAG’s adjusted suburb-level price measure rather than a simple median or property valuation. The dictionary defines suburb Median Rent as the middle weekly asking rent over a rolling 12 months; this example displays the weekly median rent returned by the market summary. It defines Gross Rental Yield as annual rent divided by Typical Price before costs and vacancy.

The snapshot captured on 10 September 2026 returned one Richmond house-market record in results[]. These are the selected fields used here:

{
  "results": [
    {
      "area_id": "VIC2167",
      "period_end": "2026-08-31T00:00:00+00:00",
      "property_type": "house",
      "bedrooms": "All",
      "typical_price": 1666826.0,
      "rent": 1054.0,
      "gross_yield": 0.0329
    }
  ],
  "total": 1
}

A careful answer would describe a typical market price of $1,666,826, weekly rent of $1,054 and gross yield of 3.29%, while keeping the selection and period visible. These are suburb-segment measures. They do not value a particular Richmond home.

Prompts that produce clearer work

Ask for the source and filters

Use HtAG data only. State the locality identifier, property type,
bedroom selection, period end and source operation beside the result.

Ask for presentation, not extra inference

Turn the returned Richmond house-market summary into a concise client note.
Do not predict future growth or infer a property valuation.

Ask for a reusable output

Format the result as a small market card and a plain-language explanation.
Treat the response as a cacheable snapshot keyed by area, property type,
bedrooms and period.

Troubleshoot from the connection outward

• No HtAG tools appear: return to the current MCP guide and verify each Claude Desktop setup step.
• Authentication fails: follow the current guide and portal instructions rather than pasting credentials into chat.
• The wrong Richmond appears: include state and postcode, then ask Claude to return the resolved locality identifier before requesting market data.
• The answer has no date: ask for period_end and the property segment explicitly.
• Claude answers from general knowledge: say “Use the HtAG tools” and require the source operation in the response.

Use the result responsibly

An MCP assistant can organise evidence and explain returned fields, but the output still needs human review. Validate the complete identity, segment and period before reuse, keep market-level figures separate from property-specific claims, and compare like-for-like periods when you add more areas.

Review the current HtAG MCP documentation for connection details or open the Developer Portal to manage access.