Case Study 2
Replacing a Rented Lead Tool With a Nextdoor System She Owns
Julia wins clients on Nextdoor, a hyperlocal platform with no API and no easy way to automate. We are building her a Claude Code system that posts, scans the feed, and surfaces leads on its own. The first piece is already live.
An Interior Designer Who Wins Work Through Her Neighborhood
Julia is an interior designer who finds many of her clients on Nextdoor, the hyperlocal network where neighbors trade recommendations. When someone in her area posts that they are looking for an interior designer, a wave of responses follows within minutes. Getting seen first, and looking credible when you do, is the whole game.
To keep up, she had been leaning on a paid third-party tool to watch the feed for her. It worked, but it was one more subscription, one more app to switch between, and a system she did not control.
Lead Generation on a Platform That Does Not Want to Be Automated
Nextdoor has no public API, so there is no clean way to plug in. Lead generation there means logging in, scrolling the feed, spotting the right posts, and responding fast, by hand, all day. For a designer running a business, that is not realistic.
What Stood in the Way
- No API to connect to, so most off-the-shelf tools do not apply
- Leads are time-sensitive: the first credible reply usually wins the job
- Posting has to stay consistent to build local credibility and recommendations
- The existing tool was a recurring cost and a black box she did not own
- Anything that reads as a bot puts the account at risk on a trust-based platform
One Owned System, Seven Workflows
We mapped the entire Nextdoor motion into seven Claude Code workflows, from posting to lead capture to follow-up. They ship in order of reliability, not flash. Here is where each one stands today.
Auto-Posting
Claude Code pulls scheduled posts from Notion, checks the date and time, and posts to Nextdoor on schedule. Playwright runs it, Chrome handles the fallback, and a notification confirms every post that goes out.
Keyword Scanner
Opens the Nextdoor feed every hour, surfaces 10 to 15 posts per pass, and checks each one against Julia's keyword list. Matches get logged to Notion with a snippet, timestamp, and review checkbox.
Keyword Filtering Layer
A whitelist and blacklist that sit on top of the scanner, so only genuine leads reach her. It cuts the noise before a single alert ever goes out.
Lead Notification
On a match, the system grabs the direct post link, logs it to Notion, and sends an instant alert so Julia can open the post and respond before anyone else does.
Auto-Comment on Matched Posts
A natural, neighborly first comment on a matched post. No sales language. Just enough to get Julia seen before the competition shows up.
Personalized Follow-Up Message
A short, personal message that reads like Julia wrote it, never a template. Personalization is what protects credibility on a platform built on local trust.
Reply Detection and Follow-Up Trigger
An hourly check for new replies and messages that pings Julia the moment a conversation starts. Since Nextdoor offers no reliable activity feed to tap into, the system scans for new activity rather than trying to parse profiles, which keeps it stable. No warm lead goes cold.
Why We Built It This Way
The interesting part of this project is not that it automates Nextdoor. It is the decisions that keep it reliable, affordable, and safe on a platform that punishes anything that feels automated.
A System She Controls
Instead of paying for a third-party monitoring tool, Julia now runs a system she controls. No subscription, no separate app to switch between, and no action layer she does not own.
Built to Drive a Real Browser
Nextdoor has no public API. Claude Code drives a real browser the way a person would, so the system works where direct integrations simply do not exist.
Hourly, Not Every Few Minutes
Local feeds are low volume, so the scan runs hourly. That is enough to beat competitors to a fresh post without hammering the platform, drawing attention, or wasting compute.
A Solid Base Before Features
The build order is deliberate. The scanner has to reach an 80 percent match-accuracy target before comment and message automation get layered on top. A reliable base first, the rest second.
Scheduled Jobs, Not Constant Calls
Routine scans run through scheduled Python jobs rather than constant AI calls, so the monitoring runs at close to zero ongoing cost. The system only spends when it is time to make a decision.
Neighbor to Neighbor, Never Spam
Every comment and message is written to read neighbor to neighbor. On a platform built on local trust, anything that feels like mass outreach does more harm than good.
Live Today, Expanding on a Reliable Base
Julia already has hands-free posting running from her Notion calendar, and the lead-detection layer is in active rollout. Every new piece is added only once the one beneath it is solid.
- Auto-posting is live: scheduled Nextdoor posts go out on their own, confirmed on every run
- The hourly feed scanner is in testing, tuning toward an 80 percent match-accuracy target
- Instant lead alerts and Notion logging are being wired directly to the scanner output
- Comment, message, and reply-detection workflows are mapped and queued next
- The whole system is owned by Julia, replacing the rented tool she used to depend on
This is what an owned automation looks like in mid-build: live where it counts, honest about what comes next, and designed to last on a platform that gives you nothing to plug into.
Have a Platform Everyone Says You Cannot Automate?
Start with a free strategy call. We will find the bottleneck, design the system, and show you exactly what we would build.