An application running on you computer would not have any referrer to send since you'd not be cross-linking from a different site as with the first scrapers you mentioned.

The best way for an application to go undetected would be for it to mimic a standard browser. In fact the application would be easily built using the web/browser frameworks in either Mac OS or Windows. It would forge its user agent to exactly match a common browser like Firefox or IE and perform the same type of normal unitary searches as CraigsHelper. The difference would be that instead of presenting you with a mess of paged results it can then locally massage the data, reformatting and presenting it to you in a completely custom fashion. Exactly what CraigsHelper claims it can't do.

You could always write a similar app in PHP and host it, but even though it would be using the cleaner searching, all the hits to craigslist would invariably come from one IP and would be extremely suspect. And blocked in short order if it became even relatively popular.

A custom application, or really custom browser UI is the way to go. Hasn't anyone done any greasemonkey or similar scripts for this kind of thing? I'd think that would be an even easier avenue and wouldn't require people to download a new app.

I don't use Craigslist primarily because you can only search a single very narrow locale at a time. Kind of pointless. That's even less useful than the old-school consulting of printed classifieds in the newspaper.