1Contact
Email herman@hermanchan.ai for a wrong or missing answer, a connection that will not authorize, a request for more calls, or a request to switch a connection off or delete its record.
2What to include
- Your connection id. Ask your AI assistant (Claude, ChatGPT or Codex) to run the connector’s data_coverage tool; its answer includes your connection_id. Every error message carries it too. It is an anonymous random id, not anything about you.
- The tool that answered (for example site_screen or rent_comps_near), and roughly when, in your time zone.
- The address, ZIP, market or name you asked about, and what you expected instead.
- The error text, if there was one.
3What to expect
You should get a reply within two business days, usually sooner. A data correction shows up in the next refresh of that layer; a service fault is fixed as soon as it is found. There is no phone or chat support.
4Known limits and coverage
- Calls: 50 successful tool calls per connection per calendar month (UTC), up to 10 a minute. Invalid input is free. Every answer shows calls_remaining_this_month.
- Supply: nine metros (Atlanta, Austin, Charlotte, Houston, Huntsville, Las Vegas, Nashville, Phoenix, Tampa). Elsewhere the supply layer says it is not covered.
- Rents: 22 rent markets, collected weekly from operators’ public listing pages, each row dated.
- Demographics and home values: any U.S. address. Outside the nine metros a demographics call can take tens of seconds.
- Not included: licensed unit counts, owner and property-manager names, and sale comps.
5Status of the data
The data_coverage tool is the live status of the data: it reports which metros hold supply rows and, for every rent market, how fresh the collection is. Ask your AI assistant to run it before trusting a thin market.