n8n automation for real estate workflows

No. And that single fact reorganizes every decision that follows. n8n's own integrations directory lists 1,953 integrations across 14 categories as of 2026, running from AI and Analytics through Sales and Utility. Real estate is not one of them. There is no MLS node, no IDX node, and no dedicated node for Follow Up Boss, kvCORE, Lofty, Sierra Interactive or BoomTown.

The vendor's own reference implementation confirms it. n8n publishes a flagship template called AI real estate agent: end-to-end ops automation, covering web-lead qualification, scheduled data research and automated voice outreach. Property data in it comes not from an MLS connector but from a generic HTTP Request node pointed at a placeholder endpoint the user supplies. The best real estate template n8n ships is a stub exactly where the real estate is supposed to be.

What you build instead

This is workable, and it is not a reason to avoid n8n. The RESO Web API is a standard OData interface, and n8n's HTTP Request node speaks it without complaint. Same for Follow Up Boss, which publishes a conventional REST API. The consequence is labor, not feasibility: you own the authentication, pagination, retry logic, rate-limit backoff and error handling a first-party node would have handled for you.

Automation landing pages present these workflows as drag-and-drop assembly: an MLS block, a CRM block. In n8n, both are the same generic node, configured by hand, twice.

Usually cheaper at volume, by a much smaller margin than comparison posts claim. Those claims fail for a specific, checkable reason: they assume Zapier bills every step in a Zap. It does not.

n8n's pricing page defines its billing unit as the entire workflow: "An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes — it's still a single execution." That is the whole basis of the cost argument, and a real advantage for long workflows. Zapier bills per successful action step, which sounds much worse until you read the exclusions. Per Zapier's own "Zapier tasks explained" help article, trigger steps, Filter, Paths, Formatter, Delay, Looping, Storage, Digest, Zapier Tables and Zapier Forms do not consume tasks. Neither do "all action steps that error or halt."

Run that against a realistic lead-routing automation: a trigger, a filter, a formatter step to normalise a phone number, a delay before the follow-up, then three app actions writing to the CRM, notifying an agent and logging the lead. Seven steps on the canvas. Three billable tasks, because the first four are all on Zapier's own exclusion list. The "20x cheaper" figure is arithmetic on a premise the vendor's own documentation contradicts.

Read the table as a shape comparison, not a price comparison. n8n prices workflow runs; Zapier and Make price the work inside them. A brokerage running a few workflows that each do a great deal will find n8n cheap. One running many two-step automations will find the difference close to noise.

Less than the sales pages suggest. NAR's 2025 REALTORS Technology Survey, drawn from 1,241 responses to 49,233 invitations (±2.78% at 95% confidence), found 46% of respondents reported AI had a neutral or no noticeable impact on their business. Thirty-three percent reported a moderately positive impact and 17% a significantly positive one.

Adoption is also concentrated in unglamorous places. Seventy-nine percent of respondents reported using eSignature and 75% social media, against 21% using a CRM with AI-powered insights and 7% using chatbots for lead capture. The industry has not quietly automated itself while you were not looking.

Lead-source data points the same direction. Asked which technology produced the highest number of quality leads in the past twelve months, 39% of respondents named social media, 23% named their CRM and 17% their local MLS. Listing syndication portals, the thing a "just listed" workflow feeds, were named by 9%. Fifteen percent named none of the options.

That 9% is a useful corrective if syndication automation tops your build list. The CRM row is highlighted because it is the one row a workflow sits inside: the failure mode there is a lead sitting unrouted, not a lead never arriving. The survey does not measure whether automating that step improves anything — it only tells you where agents say their leads come from.

One more finding reframes the pitch. Sixty-six percent of respondents said they adopt new technology primarily to save time, and only 16% cited reducing overhead or team size. The headcount-reduction story automation vendors lead with is the motivation almost nobody in the survey reported holding.

Faster than most firms manage, though the famous number quoted at you is almost certainly the wrong one. The primary source everybody gestures at is a 2011 Harvard Business Review article, "The Short Life of Online Sales Leads" by Oldroyd, McElheran and Elkington. The authors submitted test web leads to 2,241 U.S. companies and recorded the response times. The distribution is the most useful thing in the speed-to-lead genre, and no page selling you automation reproduces it.

Thirty-seven percent of the audited companies replied within an hour. Among responders, average time to first response was 42 hours. Five hundred and twelve never replied. A separate study in that same article, covering 1.25 million leads across 42 firms, associated contact within the first hour with roughly seven times the odds of qualifying a lead. Associated, not caused: firms that answer fast differ from firms that do not in many ways beyond response time.

The numbers with no source

The ubiquitous "21x more likely to qualify if you respond in five minutes" appears nowhere in that HBR article. It comes from a vendor study whose full methodology has never been published. The "7x" figure agency pages attribute vaguely to "studies" is real, but it comes from the article's secondary 1.25-million-lead analysis, not the 2,241-company audit everyone believes they are citing.

The same problem afflicts the claim anchoring most real estate automation sales pages: that agents waste 15 to 20 hours a week on administrative work. That figure is repeated across dozens of vendor pages with no attribution. We went looking for a primary source and found none. It also sits awkwardly against the closest verified datum: NAR's 2026 Member Profile puts the typical individual agent at 9 transaction sides in 2025. The vendor pages use that figure to size the transaction-administration workload automation would remove, and nine closings a year is a thin base for it. Whatever the true number is, nobody quoting it can tell you where it came from.

Treat any conversion multiplier presented without a study name, sample size and year as a marketing figure. The honest argument for fast routing does not need one: 512 companies out of 2,241 never replied at all to a lead that arrived through their own web form.

Here is the honest node path for the workflow every guide opens with. Note how much of it is not routing.

What breaks first

API rate limits, almost always on the CRM side. Follow Up Boss documents a sliding 10-second window with a 250-request global default, which sounds generous until you read the per-endpoint caps. Notes allow 10 requests per window. PUT /people allows 25. Exceed either and the API returns HTTP 429 with a Retry-After header. A naive contact-backfill loop hits that ceiling within seconds.

Scaling n8n compounds the problem rather than solving it. Running beyond a single process requires queue mode, which needs Redis as a message broker plus worker processes running a default of 10 parallel jobs each. Add workers and you multiply concurrent requests against endpoints whose own caps are as low as 10 per ten-second window. Throughput is not the bottleneck. Politeness is.

So the real work is the unglamorous part: batching requests, honoring Retry-After, idempotency keys so a replayed webhook does not create a duplicate contact, and dead-letter alerting so a failed run surfaces where a human looks. A lead that silently fails to route is worse than no automation, because the agent has stopped checking the inbox that used to catch it. These questions overlap heavily with a CRM implementation.

Two rule sets, and workflow guides routinely skip both. Start with the data. An IDX feed is licensed, not public: access requires MLS membership and data-licensing approval, usually with broker sponsorship. That is the structural reason no automation platform ships a plug-and-play MLS node.

Coverage of the modern standard is also incomplete. NAR MLS Policy Statement 7.90 has required REALTOR-owned MLSs to adopt the RESO Data Dictionary since January 1, 2016 and implement the RESO Web API since June 30, 2016. Yet as of RESO's June 2026 transition report, converted feeds cover 62.4% of U.S. subscribers and committed feeds cover 76.4%. Roughly a quarter still sit behind feeds with no stated commitment. Verify what your MLS serves before you design against OData.

On the Central Coast that has a specific answer. MLSListings, covering Monterey, San Benito, Santa Cruz, San Mateo and Santa Clara counties, states its platform updates every five minutes and serves over 6,000 brokerages across 28,000 square miles. That sets the ceiling on polling: anything faster spends executions on data that has not changed.

The listing flag that will publish something you were told not to

NAR's Multiple Listing Options for Sellers policy, effective March 25, 2025 with an implementation deadline of September 30, 2025, created "delayed marketing exempt listings": listings a seller has directed the broker to withhold from IDX and syndication for a period the local MLS sets. Any "just listed" automation must read and respect that flag, or it will eventually publish a property the seller asked you not to publish, within seconds, at scale, with your name on it. Clear Cooperation still applies alongside it: a listing must be filed with the MLS within one business day of public marketing.

Outreach: two federal deadlines, both ten business days

FCC Order 24-24 requires callers to honor do-not-call and consent-revocation requests "within a reasonable time not to exceed 10 business days of receipt," and treats a reply of "stop," "quit," "end," "revoke" or "opt out" as definitive revocation. CAN-SPAM, at 15 U.S.C. § 7704(a)(4)(A), bars commercial email more than ten business days after an opt-out request, and § 7704(a)(3)(A)(i) requires the unsubscribe mechanism to keep accepting requests for at least 30 days.

The workflow implication is precise: your cadence must check a suppression list immediately before every send, not once when a contact enters the sequence. A drip built as a linear chain of Wait nodes will happily text someone on day nine who replied "stop" on day two. Put the check inside the loop, and log it, because the log is what you produce if the question is ever asked. The same rule governs any SMS follow-up program.

Self-host only if a named person owns the server. That is the whole decision; cost is downstream of it.

The licensing question comes first because it is binary. n8n Community Edition is free under the Sustainable Use License, which permits free use, copying and modification for "internal business purposes" and non-commercial activity. Distribution is limited to sharing "free of charge for non-commercial purposes," and source files marked ".ee" require a separate Enterprise License. A brokerage automating its own operations is squarely inside internal business use. An agency reselling n8n-powered automation to other brokerages is not.

Then the operational surface. Self-hosted n8n prunes execution history by default at 336 hours (14 days) and 10,000 stored executions, because, as the documentation puts it, "your n8n database can grow in size and run out of storage." If your compliance answer depends on proving what a workflow did four months ago, the default configuration has already deleted it.

Against that, n8n Cloud Starter is €20/month for 2,500 executions and Pro is €50/month for 10,000, both billed annually. Below roughly 10,000 executions a month, Cloud will usually cost less than the hours you spend patching a VPS, and those hours are real money for an owner-operator. NAR's 2026 Member Profile puts median REALTOR gross income at $59,200 in 2025 against $9,530 in median annual business expenses. A €50 monthly line item is a visible fraction of that budget. An unbudgeted weekend recovering a Docker host is worse.

Below roughly ten transaction sides a year, workflow automation is a hobby rather than an operating advantage. NAR's 2026 Member Profile puts the typical individual agent at 9 sides in 2025 against a median of 32 for teams, with 21% of members working on a team and average team size of four. That is a volume threshold, not a published finding: in our own builds, a coordination workflow at single-digit annual sides has cost more to build and maintain than it returned, and the case improves roughly in step with transaction count.

Volume is the qualifying variable, which is why team builds succeed where solo builds stall: a team has the transaction count to amortise the build and someone whose job includes noticing when it breaks. Market conditions set the pacing. C.A.R.'s May 2026 Home Sales and Price Report put the Central Coast median existing single-family price at $1,126,250, essentially flat year over year at +0.1%, against a California statewide median of $930,260. Sales moved instead: Central Coast sales led all regions with a 21.4% year-over-year increase, Monterey County up 41.2% at 15.5 median days on market and San Benito up 30.0% at 17 days. Where price is flat and volume climbing, the constraint is throughput per transaction. Those medians of roughly two to two and a half weeks also set the runway a coordination cadence works within.

What it cannot do

n8n cannot do anything requiring a license. It cannot advise, sign, or interpret a disclosure. It will not replace a transaction coordinator, because most of that role is judgment and chasing humans. What it replaces is the part of the job that is a checklist with dates attached: deadline reminders, document-status polling, notification fan-out, CRM hygiene. Scope it to that and it holds up.

And hold the survey finding in view while you scope. Forty-six percent of REALTORS reported AI had a neutral or no noticeable impact on their business, and the motivation they did report was time, not headcount: 66% adopt technology to save time, 16% to reduce overhead or team size. Scope to the first goal and you are building toward something respondents say they want. Scope to the second and you are chasing an outcome the industry's own survey does not show many people reporting. Our marketing for real estate and AI automation hubs cover where these systems sit in a full stack.