Scout and underwrite real estate deals nightly

A builder made a Claude Code agent that scans hundreds of MLS listings nightly, underwrites those passing filters, and emails qualifying deals to investors before 6am. · Jul 7, 2026

Done withCCClaude Code

What they did

The builder says the agent runs on a nightly schedule with Inngest, scans hundreds of MLS listings, filters by price and cap rate, then underwrites the survivors. It pulls listing data from a property-data API, backfills missing beds, baths, and square footage from a secondary endpoint, runs multiple AI models in parallel to synthesize a score, and uses user actions like saves, passes, and closes to tune fit scoring over time.

How it went

They say it became Dealsletter, with almost 200 users and over 2,400 newsletter subscribers. The main shortfalls were sparse listing data, cold-start learning, and duplicate properties across feeds, which required confidence scoring and reconciliation.

Worth knowing

The biggest catch was data quality rather than scheduling: missing multifamily fields forced a backfill layer, and duplicate listings across sources needed extra reconciliation logic.

Source: Reddit · Jul 7, 2026