crewai-zillapi

CrewAI tools for Zillapi, a REST API for Zillow property data.

Give a CrewAI agent the ability to look up a property by address, zpid or Zillow URL, pull valuations, price history, tax history, schools, photos and comparable homes, and search for sale, for rent or sold listings inside a geographic area.

Install

pip install crewai-zillapi

Get an API key

Create a key at zillapi.com. You get 100 credits at signup, no card, one time. Paid plans start at $5 per month. See pricing.

Set the key in your environment:

export ZILLAPI_KEY="zk_your_key_here"

Quick start

from crewai import Agent
from crewai_zillapi import (
    ZillapiPropertyByAddressTool,
    ZillapiPropertyDetailTool,
    ZillapiListingSearchTool,
)

analyst = Agent(
    role="Real estate analyst",
    goal="Answer questions about properties and local markets with real data",
    backstory="You research homes and neighborhoods using live Zillow data.",
    tools=[
        ZillapiPropertyByAddressTool(),
        ZillapiPropertyDetailTool(),
        ZillapiListingSearchTool(),
    ],
)

Use a single tool directly:

import json
from crewai_zillapi import ZillapiPropertyByAddressTool

result = json.loads(
    ZillapiPropertyByAddressTool().run(address="350 5th Ave, New York NY 10118")
)
print(result["data"]["zpid"])

Tools

Failed calls are never charged.

ZillapiPropertyDetailTool sections: zestimate, price-history, tax-history, schools, photos, agent, nearby, open-houses, facts.

Searching an area

A listing search needs a real bounding box in decimal degrees. A place name on its own is not accepted, so convert the place to coordinates first.

import json
from crewai_zillapi import ZillapiListingSearchTool

result = json.loads(
    ZillapiListingSearchTool().run(
        west=-73.99, south=40.74, east=-73.98, north=40.75,
        status="for_sale", beds_min=2, price_max=2_000_000, max_items=25,
    )
)

Because a search costs 1 credit per listing returned, keep max_items tight. Asking for more than 50 listings runs the search asynchronously and returns a job_id, which you poll with ZillapiJobTool until its status is succeeded, then read with action="results".

Response shape

Every tool returns a JSON string with a stable envelope, and the tools never raise from execution. An agent can always parse the result.

{
  "success": true,
  "data": { "zpid": "1234567", "price": 1250000 },
  "request_id": "req_abc123"
}

{ "success": false, "error": { "code": "invalid_filters", "message": "..." } }

data is the payload itself, so you never have to reach through a second wrapper. request_id is worth logging if you need to ask about a specific call. List endpoints also return a meta object alongside data.

Match on code, never on message.

Coverage

These tools cover the synchronous, agent friendly surface of the API. The full REST API also offers batch property jobs, chained search with details, building unit extraction and outbound webhooks. See the API documentation and the OpenAPI spec.

Zillapi is also available as an MCP server for MCP compatible clients.

Notes

Zillapi is an independent service and is not affiliated with, endorsed by, or sponsored by Zillow, Inc. Zillow is a trademark of its respective owner. Use of the data is subject to the Zillapi terms.

License

MIT