Skip to content

A quieter way to search

Search Luxembourg homes from ChatGPT or Claude

Bring live listings, market data and source context into the assistant you already use.

ChatGPT

Custom MCP apps require developer mode; availability depends on your ChatGPT plan and workspace permissions.

Claude

Add the remote connector under Customize → Connectors; organization owners may need to add it first.

MCP

Add the ChatHome connector from your assistant's integrations. Availability and menu names vary by client and plan.

https://chathome.lu/api/mcp

Connect your assistant

Public property search works without a ChatHome account. Higher limits and private actions require sign-in.

Use Settings → Connectors to add this URL in Claude web or Desktop. The optional file below uses the third-party mcp-remote bridge and needs Node.js; a remote URL alone does not work in this file.

https://chathome.lu/api/mcp
claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "chathome": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://chathome.lu/api/mcp"
      ]
    }
  }
}

Setup guide

Where the data comes from

ChatHome builds its own index from the public websites of estate agencies in Luxembourg and its cross-border commuter belt. Each listing names its agency, links the agency's original page and says when ChatHome last saw it there; a fact the agency page does not state stays unknown. Agencies can ask to have their listings removed. Market figures come from public statistics, named in each reply. All data sources

MCP Server · Streamable HTTP · 2025-11-25

chathome MCP

Connect Claude, Cursor, or ChatGPT to live listings in Luxembourg and its cross-border commuter towns, plus Luxembourg market intelligence. Search rentals, browse listings for sale, and pull commune data straight from your AI assistant.

MCP Server URL

https://chathome.lu/api/mcp

Authentication

Every tool works without an account; connect one only for higher limits.

Start immediately — no login

Every tool runs without an account: search, listing details with the agency's contact, market data, estimates, calculators and commute times. Anonymous calls use smaller hourly limits.

AOAuth 2.1 with PKCERecommended

Your MCP client handles the OAuth flow automatically — no Client ID or secret to enter. This server uses public clients + PKCE with dynamic client registration (RFC 7591): your client registers itself the first time it connects.

Claude Code — project configuration in .mcp.json

json
{
  "mcpServers": {
    "chathome": {
      "type": "http",
      "url": "https://chathome.lu/api/mcp"
    }
  }
}

How the flow works

  1. 1Paste the MCP URL into your AI client — no Client ID or secret needed.
  2. 2The client opens a browser to chathome.lu where you log in and approve access.
  3. 3Tokens are issued automatically — the client stores them and refreshes as needed.
or
BAPI KeyFor automation / scripts

Generate a key at Dashboard → Agent Access, then pass it as an X-API-Key header. Keys are prefixed cht_.

json
{
  "mcpServers": {
    "chathome": {
      "type": "http",
      "url": "https://chathome.lu/api/mcp",
      "headers": {
        "X-API-Key": "cht_YOUR_KEY_HERE"
      }
    }
  }
}

Rate limits

  • ·200 tool calls / hour per user
  • ·300 calls / hour per IP
  • ·Tokens expire after 1 hour (OAuth auto-refreshes)

Available Tools

21 tools covering rental search, for-sale search, listing detail, market data, price estimation and commute times.

Response Shapes

Each tool returns a typed JSON object. Search tools wrap listings in an array.

search_rentals · search_for_sale

typescript
{
  listings: AgentListing[];
  total: number;                 // exactly listings.length
  returnedCount: number;
  matchingCount: number | null;
  matchingCountKind: "exact" | "lower_bound" | "unavailable";
  marketMatchingUpperBound: number | null;
  hasMore: boolean;
  nextCursor: null;              // no stable server cursor yet
  searchExhaustive: boolean;
  truncated?: true;
}

get_listing

typescript
{
  listing: AgentListing;                 // normalized decision record
  listingPage: PublicListingProjection; // complete fail-closed page data
  listingPageContext: ListingPageContext; // price, area, amenities, passport, parcel, transport
  trust: TrustEnvelope;
}

AgentListing type

typescript
type AgentListing = {
  schemaVersion: "agent-listing.v2";
  id: string;
  url: string;
  title: string;
  status: "active";
  price: number | null;
  currency: "EUR";
  listingType: "rent" | "sale";
  propertyType: string | null;
  city: string | null;
  district: string | null;
  country: "LU" | "FR" | "BE" | "DE" | null;
  beds: number | null;
  baths: number | null;
  areaSqm: number | null;
  energyClass: string | null;
  imageUrl: string | null;
  agencyName: string | null;
  source: {
    type: "original_agency_listing";
    agencyName: string | null;
    listingUrl: string;
    agencyListingId: string | null;
    // Latest catalogue re-sighting, not a detail-page fetch claim.
    lastSeenAt: string | null;
    status: "active" | "under_offer" | "unknown";
    statusBasis: "source_catalog_seen" | "listing_record_active";
    agencyConfirmedAt: string | null;
  } | null;
  rent: {
    baseMonthly: number | null; chargesMonthly: number | null;
    monthlyTotal: number | null; monthlyTotalStatus: string | null;
    knownMonthlyLowerBound: number | null;
    estimatedMonthlyTotal: number | null;
    estimatedUtilitiesMonthly: number | null;
    deposit: number | null; depositMaximum: number | null;
    agencyFee: number | null; agencyFeeBasis: string | null;
    mandatoryRecurringFees: MandatoryRentalCost[];
    mandatoryMoveInFees: MandatoryRentalCost[];
    knownUpfrontCost: number | null; exactUpfrontCost: number | null;
    unknownComponents: string[];
  } | null;
  availability: { offerStatus: "available" | "under_offer" | "unknown";
    availableFrom: string | null; moveInState: string | null };
  property: { rooms: number | null; floor: number | null;
    constructionYear: number | null; heatingType: string | null;
    hasFloorplan: true | null };
  location: { city: string | null; district: string | null;
    address: string | null; postalCode: string | null;
    precision: "exact_address" | "street" | "district" | "commune_only" | "unknown" };
  history: { firstObservedAt: string | null; daysListed: number | null;
    priceDropTotalEur: number | null;
    lastChangedAt: string | null };
  duplicates: { confirmedOtherSources: number; confirmedSourceCount: number };
  officialScreening: OfficialScreening | null;
  propertyChecks: ExactPropertyChecks | null; // all 7 checks, incl. noise
  transport: PropertyOrCommuneTransport | null; // station service + commute hubs
  valueContext: { priceIntegrity: object | null; fairPrice: object | null;
    marketContext: object | null };
  fieldProvenance: Record<string, { publisher: string | null;
    method: string; sourceUrl: string | null; observedAt: string | null }>;
  missingFields: string[];
  match: { allRequiredConstraintsSatisfied: boolean | null;
    requiredConstraints: Array<{ id: string; status: string }>;
    reasons: string[] };
  latitude: number | null;
  longitude: number | null;
  // True when at least one factual field has source evidence.
  hasSourceCheckedFacts?: boolean;
};

get_market_context · get_price_estimate · get_commute_times

typescript
// Each returns a tool-specific JSON object, e.g.:
// get_price_estimate → { estimatedPrice, low, high, confidence, comparables, … }
// get_commute_times  → { commune, destination, mode, durationMinutes, … }
// get_market_context → { commune, dimensions, lastUpdated, sources }

Resources & Prompts

The server also exposes one resource template and two built-in prompts.

Resource

chathome://listing/{id}

Read a single listing by ID — same data as get_listing, exposed as an MCP resource URI.

Prompt · find-family-home

Guided search for family-friendly rental properties. Arguments: commune?, budget?, beds?.

Prompt · match-properties

Natural-language property search followed by exact listing cards. Arguments: criteria, intent?, locale?.

Try It

Once connected, ask your AI assistant:

  • “Find me 2-bedroom apartments in Luxembourg-Ville under €2,000/month”
  • “What are houses for sale in Strassen with a garden?”
  • “Compare these two listing IDs using their source-backed facts”
  • “Show me the full details for listing ID abc-123”

Claude Agent Skill

A packaged skill that teaches Claude how to use chathome well — tool selection, Luxembourg and commuter-belt search workflows, and Luxembourg market context.

claude.ai & Claude Desktop

Download the zip, then upload it under Settings → Capabilities → Skills. Pair it with this MCP server added as a custom connector.

Download skill (.zip)

Claude Code — one-step plugin (skill + MCP server)

bash
/plugin marketplace add chathome-lu/chathome-claude-plugin
/plugin install chathome@chathome

Agents & other runtimes

The skill follows the open Agent Skills standard. Fetch it raw at /skills/chathome/SKILL.md.

Roadmap

What's live now and what's coming next.

Live

MCP v1 — 21 tools, 1 resource, 2 prompts over Streamable HTTP; every tool works anonymously, and OAuth 2.1 + PKCE or an API key gives higher limits.

Live

A2A Agent Card — /.well-known/agent-card.json for Google A2A protocol discovery (task endpoint /api/a2a).

Live

Agent Skill — downloadable Claude skill bundling chathome workflows, paired with this MCP server.