An AI real estate skill that finds homes worth a closer look.
Search US homes with ZillAPI MCP, compare prices, and get a ranked photo shortlist. Tell your agent what you like. Let it remember your preferences and repeat the search on your schedule.
Install this skill and onboard me:
https://github.com/ZeroPointRepo/property-scanner
Add the Grok Bot → · Install with npx · How it finds opportunities
Austin, under $475k, at least 3 bedrooms. I'd prefer a home without a pool.
The agent walks you through connecting ZillAPI with OAuth, saves your search, and gets to work. Its shortlist gives you:
- A photo and the basics: address, asking price, bedrooms, bathrooms, and size.
- A reason to look: the price gap, supporting comparisons, and relevant price cuts.
- The evidence behind the number: how many comparisons were used, whether prices are actual sales or pending asking prices, and what needs review.
- An easy next step: reply with a home's number to see its comparisons and listing details.
“Exclude fixer-uppers” changes future searches. “Run this every weekday morning” sets up a routine in an agent with scheduling support. Recurring updates focus on new matches and meaningful changes.
Prefer a ready-made bot? Open Property Scanner in Grok Bot and follow its onboarding.
The template runs in Grok Bot. This repository contains the original portable skill for other compatible agents.
With Node.js installed, run:
npx skills add ZeroPointRepo/property-scanner --skill property-scannerChoose your agent in the installer, then ask it:
Use Property Scanner and onboard me.
The Skills CLI installs the skill, references, and calculation helper together. Your agent needs Agent Skills support and access to ZillAPI MCP. Persistent memory enables saved preferences; a scheduler enables routines. The included helper uses Python 3 and its standard library; the skill also specifies the calculations for hosts with another execution tool.
For a manual installation, copy the entire skills/property-scanner directory into your agent's skill location. Keep its references and scripts folders with SKILL.md.
Property Scanner compares an asking price with nearby property evidence and Zestimate. It starts in relaxed mode to surface leads for review; strict mode requires stronger comparison coverage.
| Relaxed · default | Strict · optional | |
|---|---|---|
| Minimum price gap | 3% | 5% |
| Comparison homes | At least 1; prefer 2+ | At least 3 |
| Search radius around a home | 2.5 miles | 1 mile |
| Missing Zestimate | Use comparison reference; flag for review | Cannot assess |
| Missing closed sale price | May use a recent pending asking price; flag for review | Exclude that comparison |
The calculation uses median comparison price per square foot × the home's size. When Zestimate is present, the lower reference determines the gap. Actual sales take precedence over asking-price substitutes for the same sale episode. Any substitute is labeled; a mixed set is never presented as pure sold-price evidence.
The bot also checks similarity, listing concerns, and your preferences. A price cut alone does not qualify a home. See the full comparison and ranking rules.
These are leads to investigate, not appraisals or guaranteed bargains. One comparison, asking-price substitutes, missing estimates, and substantial property differences can weaken the signal. The bot keeps that context visible.
ZillAPI charges per returned record. The skill limits fetching and reuses saved comparisons:
| Work | Default credit ceiling, including lookups |
|---|---|
| Active scan · up to 10 listings | 20 |
| Comparison refresh · up to 20 listings, initially and at most weekly | 40 |
| Initial scan and comparison setup | 60 |
| All scanner activity per billing period | 900 |
These are instructed budgets, not measured usage guarantees or a limit enforced by ZillAPI. The agent tracks usage and stops before exceeding them. Manual scans and follow-ups share the allowance. Smaller searches provide focused coverage; they do not inspect every home in a city.
You need your own ZillAPI account and sufficient credits. OAuth is preferred; API keys are a fallback. Check current ZillAPI pricing and the fetching and budget rules.
The entry point stays short. The agent loads details for its current task:
| File | Job |
|---|---|
| SKILL.md | Core behavior and routing |
| Onboarding | Connect, establish preferences, and run once |
| ZillAPI | Data fetching, OAuth, and credit limits |
| Scanning | Matching, formulas, and ranking |
| Delivery | Photo messages and numbered follow-ups |
| Preferences and routines | Learn from feedback and repeat the search |
The calculation helper works on normalized, prefiltered records. It makes no network calls. Property retrieval, semantic matching, authorization, messaging, and scheduling are handled by the host agent.
To run the calculation tests:
python3 -m unittest discover -s tests -vFound a bug or have a useful improvement? Open an issue. Use an anonymized example without credentials or personal search history.
Powered by ZillAPI. MIT licensed.
ZillAPI is an independent service, not affiliated with Zillow Group. Zillow and Zestimate are trademarks of Zillow Group. The Grok Bot template is user-created; this repository is not an official xAI product.