Score off-market property leads from inboxes

They used Claude and Python to pull deal details from email, score the listings, and route the results into tools the team could review.

Done with Claude CodeClaudeยท Intermediate

What they did
They created a dedicated Gmail account to collect saved-search notices and wholesaler messages. A Python job then separated emails that included more than one property, sent each property through Claude to fill a fixed JSON structure with blanks left as null, and let ordinary Python calculate scores for single-family and multifamily deals. Costly enrichment calls only ran after a cheap price check passed. The finished records were published to Supabase, displayed in a Streamlit dashboard, surfaced in email alerts, and copied into a shared Google Doc for coordination.
How it went
This setup was enough to gather incoming leads in one place and score the ones with enough data. Many records still stopped at a needs-info stage because about half came through without a full location, and others arrived missing key rental details like unit totals or rents. Login-restricted sources were another reason some deals could not be processed automatically.
Tips
Put a cheap screen in front of any paid lookup so you only spend on deals worth reviewing. Keep missing fields as null and send those entries into a follow-up queue instead of trying to guess.

Try it yourself with Claude Code

Help me build a Python workflow that monitors [mailbox or .eml files] for off-market real estate emails, splits messages that contain several deals, and asks Claude to extract each one into a fixed JSON schema with fields such as [location, list_price, unit_count, rent, NOI, seller_contact]. If any field is unavailable, set it to null and add that deal to a follow-up list. After that, use plain Python to score each record against [my rules]. Pause and ask for confirmation before any irreversible action like paying for enrichment, sending notifications, deleting messages, or submitting forms. I want to test it on [N] sample emails and get a ranked output with scores and missing-data flags.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on reddit.com โ†—

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.