MCP Server

    Homecastr for AI Agents

    Plug Homecastr into any MCP-compatible agent. Your AI gets real probabilistic forecasts — not hallucinations.

    Install in 2 minutesREST API docs

    Desktop clients

    Native MCP support in desktop assistants. Forecasts appear inline in conversation.

    Editor clients

    Drop into your editor's MCP config and your coding assistant can reason about home values.

    Any MCP client

    Standard Model Context Protocol — works with any client that supports MCP 2024-11-05+.

    Installation

    1. Install the server

    bash
    # Install Homecastr MCP server
npm install -g homecastr-mcp-server

# Or run directly with npx (no install)
npx homecastr-mcp-server

    2. Add to your MCP client

    Edit ~/.config/homecastr/mcp.json:

    json
    {
  "mcpServers": {
    "homecastr": {
      "command": "node",
      "args": ["/path/to/homecastr-mcp-server/dist/index.js"],
      "env": {
        "HOMECASTR_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

    Or add to Cursor

    json
    // .cursor/mcp.json
{
  "mcpServers": {
    "homecastr": {
      "command": "node",
      "args": ["/path/to/homecastr-mcp-server/dist/index.js"],
      "env": { "HOMECASTR_API_KEY": "YOUR_API_KEY" }
    }
  }
}

    3. Get your API key

    Free with email confirmation, no sales call. Get your key →

    Example Conversation

    Homecastr MCP
    User: What's the 5-year forecast for 1000 Brickell Ave Miami FL?

Assistant: Let me look that up for you.
[calls forecast_address("1000 Brickell Ave Miami FL", 2030)]

Based on Homecastr's model, 1000 Brickell Ave has a median forecast of $425,000
by 2030 (+12.5% appreciation). The confidence range is $385K–$480K.

    Available Tools

    All tools are called automatically by the AI — you just describe what you want in plain English.

    forecast_address

    Forecast home values by US street address. Geocodes automatically.

    Parameters

    address*stringUS street address (e.g. "1000 Brickell Ave Miami FL")
    yearintegerForecast year 2026–2030 (default: 2030)

    Returns

    P10/P50/P90 forecast bands, appreciation %, multi-year fan chart, H3 cell ID

    forecast_hex

    Forecast by H3 cell ID for neighborhood-level analysis.

    Parameters

    h3_id*stringH3 cell ID at resolution 8 (e.g. 882a100c65fffff)
    yearintegerForecast year (default: 2026)

    Returns

    Same as forecast_address

    forecast_parcel

    Forecast by tax parcel account ID. Supports Houston HCAD, NYC, Florida, Seattle.

    Parameters

    acct*stringCounty tax account / parcel ID
    yearintegerForecast year

    Returns

    Parcel-level P10/P50/P90 forecast

    get_market_overview

    Aggregate forecast for a market — county, state, or ZCTA.

    Parameters

    geoid*stringFIPS code or ZCTA (e.g. 48201 for Harris County)
    yearintegerForecast year

    Returns

    Median forecast, appreciation, property count, geographic metadata

    compare_markets

    Compare forecasts across multiple locations side by side.

    Parameters

    locations*string[]Array of addresses or GEOIDs to compare
    yearintegerForecast year

    Returns

    Ranked comparison table with appreciation % and absolute forecast values

    Common Use Cases

    Real Estate Research

    • ›“What's the 5-year outlook for homes in Harris County?”
    • ›“Compare appreciation in Houston vs NYC vs Chicago”
    • ›“Find the top 10 tracts with >15% projected growth”

    Investment Analysis

    • ›“Evaluate the risk profile for parcel 00404328020003500”
    • ›“What's the downside scenario (P10) for this zip code?”
    • ›“Rank these 5 properties by expected appreciation”

    Underwriting

    • ›“What's the 2028 forecast for 1000 Brickell Ave?”
    • ›“How confident is this forecast? What's the band width?”
    • ›“Pull forecasts for this list of 50 parcels”

    Market Reports

    • ›“Summarize the Texas housing market outlook to 2030”
    • ›“Which states have the strongest appreciation outlook?”
    • ›“Build a comparative table for Sun Belt markets”

    Data Coverage

    GeographyCoverageScale
    Parcel / UnitHouston, NYC, Chicago, Seattle, Florida46M+ units
    Census BlockNYC + Houston67K blocks
    Census TractFL, Houston, nationwide (ACS)82K tracts
    ZIP Code (ZCTA)Nationwide20K ZCTAs
    CountyNationwide3,230 counties
    StateAll 50 states + DC51 states

    Troubleshooting

    MCP server shows HTTP 404 in Status page▾
    Restart the server: kill the homecastr-mcp process, then re-add the server config in your MCP client settings.
    Geocoding failed for my address▾
    Use format "Street City STATE" (e.g. "1000 Brickell Ave Miami FL"). Avoid ZIP codes, abbreviations like "St." instead of "Street", or apartment numbers.
    No forecast data for my location▾
    Coverage is currently strongest in Houston, NYC, Chicago, Seattle, and Florida. Nationwide tract/county/state forecasts are available everywhere.
    How often does data update?▾
    Forecasts refresh weekly (Tuesdays ~2 AM UTC). Check /status for current data freshness.

    Ready to give your AI real data?

    Get a free API key and have your agent forecasting in under 2 minutes.

    Get API Key — FreeREST API Docs