Build Your Own AI Sourcing Engine

Most sourcing tools surface 100 'perfect matches' and 80 are not close. The judgment that makes you good (seniority calibration, which adjacent companies translate, tenure red flags, the candidate type your client always says yes to) never makes it into the tool. This is how Crustdata customers build their own sourcing engine on Claude over MCP, with that judgment baked in.

The workflow, step by step

1. 1
   Define the brief as filtersTranslate the role into People Search filters: current_employers.title, current_employers.seniority_level, skills, and location.
2. 2
   Fetch a wide poolPull a large candidate set first. Contact data stays untouched, so a wide net is cheap.
3. 3
   Let the agent apply judgmentHave the agent rank on adjacent-company fit, tenure, and your client's past yeses, before any enrichment.
4. 4
   Enrich only the finalistsWaterfall-enrich the shortlist for verified emails and phones.