Your second brain
for closing deals.
Speak after a showing. Forward an email. Pull up a client. Lumi captures the soft signals, fills the brief, and feeds Claude — automatically.
- Voice → CRM, auto. No forms.
- Works offline. Syncs when you're back.
- Free for agents in EU · LatAm · MENA.
9-min read · Updated April 2026
Lumi · Wednesday
Good morning, Niki.
Two showings · three leads need a nudge.
Showing · Passeig de Gràcia 84
The 2-minute lookup
that tells you everything.
A first call with a buyer is where most agents calibrate badly — too aggressive on price, too soft on timeline, too generic on the shortlist. The agents who calibrate right walk into the call already knowing how this person transacted last time. Public records told them. They just had to look.
PRIOR PATTERNS
· 2 purchases, 2 sales over 10y
· medium hold (3.4y, 5.2y)
· 2016 -5% asking, 2019 at asking,
2024 sold +7% market
OWNERSHIP STYLE
· solo all 4 transactions
· permits show post-purchase
renovation pattern
ASK EARLY:
"Settling-in purchase, or option you'd
revisit in 2-3 years?"Two minutes of records pull. Eight seconds of reading. The first call runs differently because the agent walks in calibrated.
What 30 minutes on the phone won't tell you.
On a first call, the questions agents ask reveal what they don't know. “Have you bought before?” “Solo or with a partner?” “Any timeline pressure?” Each question is information the agent needs — and each one is information already in the public record. Asking signals that the agent didn't prepare. The buyer notices.
The records-first protocol inverts this. The agent walks in knowing the prior transactions, the typical hold time, whether this buyer tends to negotiate hard or pay at asking. They don't lead with that knowledge — they don't mention it at all — but they don't waste the call discovering it either. They use the call for what it's actually for: feeling out fit, testing rapport, surfacing the soft signals the records can't show.
The output of the call is a relationship that's 30 minutes ahead of where it would have been otherwise. The patterns are already in the agent's head. The next move is calibrated. Three or four calls in, the agent has compounded this advantage into a deal that the un-prepared version would have lost.
“The questions agents ask reveal what they don't know. And what they don't know is mostly in the public record.”
What to query. By region.
Each source is publicly accessible in every market Lumi serves (EU, LatAm, MENA), with varying levels of programmatic API. The names and access methods differ by jurisdiction; the data shape is consistent.
Every prior purchase and sale by name. Reveals hold time, transaction frequency, price patterns vs market.
Portugal: Conservatória do Registo Predial. Spain: Registro de la Propiedad. UAE: DLD. Brazil: Cartório de Registro de Imóveis. Most accessible online with a small fee per query.
Renovation history, building permits, declared works value. Reveals whether they buy turnkey or invest post-purchase.
Portugal: Câmara Municipal. Spain: Ayuntamiento. UAE: Dubai Municipality. Brazil: Prefeitura. Often online, varies by city.
Any litigation involving property — disputes, liens, foreclosures, contested sales. Major red flag if any.
Portugal: Citius (free). Spain: Boletín Oficial del Estado. UAE: Court records (DIFC, ADGM). Most LatAm jurisdictions: regional court portals.
LLC, trust, or company ownership patterns. Tells you whether they hold for tax/estate reasons or personal use.
Portugal: Registo Comercial. Spain: Registro Mercantil. UAE: free zones (DIFC, ADGM, JAFZA). Brazil: Junta Comercial.
Five rules. Two minutes per buyer.
The discipline of records-first preparation. Skip any one and the protocol either burns hours per lead or produces a creepy first conversation.
- 01
Run the cross-reference before the first call. Not after.
The whole value of this protocol is in the strategy it informs in the first conversation. After the first call, you've already calibrated wrong on price talk, timeline questions, and decision-group probes. Pull records the day the lead lands; review the dossier the morning of the first call. Ten extra minutes; the rest of the relationship runs differently because of it.
- 02
Four sources or it's not a cross-reference.
Pulling deed registry alone gives you transactions but not the why. Permits without deeds gives you renovation activity without context. The four-source pull (deeds + permits + courts + corporate registry) is what produces a dossier that holds together — each source corroborates or contradicts the others. One source is data; four sources are intelligence.
- 03
Claude compresses 4 sources into 4 sentences.
The records output is messy — different formats per registry, prices in different currencies, dates in different conventions. Use Sonnet to normalise and synthesise; the 4-sentence schema enforces compression. The agent reads the 4 sentences in 8 seconds and has the calibration. The full source data is filed for compliance and never re-read.
- 04
The dossier is for the agent's strategy. Never the client's eyes.
Crucial discipline: the public-records dossier informs how the agent runs the relationship — what questions to ask early, what price-calibration to test, what timeline to assume — but never appears in any client-facing artefact. Mentioning the records explicitly to the client ("I see you bought your last property under asking") feels like a hostile move and breaks the warmth. The records work invisibly.
- 05
If you find litigation or liens, treat as a red flag for due diligence.
Court filings is the only one of the four sources that can produce material risk findings — a lis pendens, a tax lien, a contested estate. If found, the protocol switches from "calibrate the relationship" to "surface it through the agent's normal due-diligence disclosure to lender / lawyer / counterparty". Don't bury it; don't lead with it; route it through the legitimate channel.
Three failure modes that burn the relationship.
Each of these has been done by a real agent we worked with. Each one cost them either the deal in front of them or the next 18 months of repeat business.
“Marina, I noticed in the deed registry that you bought your Lisbon flat under asking in 2016 — clearly you know how to negotiate. We can use that experience for your Porto purchase!”
Mentions the records to the client. Even though everything is public, this feels like surveillance — and erodes the warmth the agent needs to do the rest of the relationship. The dossier is for the agent's eyes only.
“Based on the name change in 2018 from Costa to Costa-Almeida and back to Costa in 2022, Marina likely went through a divorce — she may be price-sensitive and emotionally fragile around the new purchase.”
Inferring marriage / divorce from name changes is wildly unreliable (could be professional alias, surname order convention, marriage-then-reverted-by-choice, half a dozen other reasons). The dossier should report records facts, never psychological inferences.
“Based on Marina's transaction patterns, I think she will: 1) want to negotiate hard on price, 2) prefer a renovation project, 3) make the decision alone. Plan accordingly.”
Pre-baking conclusions denies the agent the chance to test them. The dossier surfaces patterns; the agent draws the conclusions in real time on the first call. AI conclusions get over-trusted; AI patterns get verified.
The cross-reference input.
What the prompt receives after pulling all four sources. The agent's note guides the synthesis but doesn't override it — the records are what they are.
# ── public records cross-reference ──────────────
subject: "Marina Costa"
agent_note: "First call tomorrow.
Want to know transaction style."
deed_registry (Lisbon + Porto):
- 2016-03: purchased Lisbon 1-bed,
€280k (asking €295k, sold under)
- 2019-08: sold Lisbon 1-bed, €340k
(held 3.4 years)
- 2019-09: purchased Lisbon 2-bed,
€420k (asking €420k, at asking)
- 2024-11: sold Lisbon 2-bed, €510k
(held 5.2 years)
- 2024-12: signed lease Porto, no
purchase yet
permit_office (last 10 years):
- 2017: kitchen renovation Lisbon 1-bed
(€18k declared)
- 2021: bathroom + balcony Lisbon 2-bed
(€34k declared)
court_filings: none in last 10 years
corporate_registry: no LLC ownership;
all purchases solo
co_buyer_pattern: solo across all
name_change_history: none
inferred_market_position:
2016 purchase: undermarket by ~5%
2019 purchase: at asking (competitive
market that year)
2024 sale: overmarket by ~7%
(Lisbon peak, sold well)
What to feed Claude.
The system prompt that turns four messy public-records pulls into the 4-sentence dossier. Sonnet recommended for the synthesis nuance — Haiku tends to over-conclude on thin data.
You are a senior real-estate agent's
public-records analyst.
INPUT
You receive structured data from 4 public
sources: deed registry (prior purchases and
sales), permit office (renovation history),
court filings (any property-related disputes
or liens), and corporate registry (LLC or
trust ownership patterns). Plus the agent's
note on what they're trying to learn.
OUTPUT
A 4-sentence dossier — same shape as the
lead-enrichment dossier — but focused on
prior real-estate behaviour:
1. PRIOR PATTERNS — how many transactions
this person has done, their typical
hold time, whether they tend to buy
at asking / over / under.
2. OWNERSHIP STYLE — solo, joint, LLC,
trust; whether co-buyers have changed
across transactions; any pattern that
suggests a financial-planning approach.
3. NEGOTIATION SIGNAL — based on prior
transaction prices vs market at the
time, do they tend to over-pay, under-
pay, or close at expected price.
4. WHAT TO ASK — one specific question
the agent should ask early in the
first conversation, calibrated to the
prior patterns above.
RULES (non-negotiable)
1. Public records only. If a field is
blank or unverifiable, skip it — never
guess or infer beyond what the records
show.
2. Never reference the records to the
client. The dossier informs the
agent's strategy; it never appears in
any client-facing message.
3. The fourth sentence is calibrated to
the patterns. If they hold long, ask
about timeline. If they tend to under-
pay, ask about decision-criteria.
4. No private records, no breach data,
no people-search dossiers. Sources are
the deed registry, permit office,
court records, corporate registry.
5. Voice: clinical-detective. Strategic,
not gossipy.
ANTI-PATTERNS (never produce these)
- "Based on the records, I think they
will…" (this is for the agent to
conclude; the dossier reports facts)
- Any reference to the records in the
"what to ask" sentence
- Speculation about marriage / divorce
from name changes (often misleading)
- Including data older than 15 years
unless it shows a clear pattern.
The agent should read the 4 sentences in
8 seconds and walk into the first call
already knowing the patterns.Pull the records the day the lead lands; run the prompt the morning of the first call.
What Claude returns.
The dossier that the agent reads in 8 seconds before the first call. Notice that the fourth sentence is a question — calibrated to what the records reveal about Marina's transaction patterns.
1. Marina has done 2 prior purchases and 2 prior sales over 10 years, with hold times of 3.4 and 5.2 years — she's a medium-term buyer, not a flipper or a forever-home holder, suggesting she'll evaluate the Porto purchase against a 4-7 year horizon. 2. All four prior transactions were solo (no co-buyer or LLC), and her permit history shows she invests in renovations during ownership rather than buying turnkey — she'll value renovation potential over move-in-ready, and she'll handle the decision herself rather than deferring to a partner or advisor. 3. She closed her 2016 purchase ~5% under asking and sold her 2024 property ~7% over market — she's a disciplined buyer who walks away from inflated asks, but she's not afraid to negotiate hard when she identifies a property she actually wants. 4. Ask early: "What kind of timeline are you working against — is this a settling-in purchase or an option you'd give yourself two-three years to revisit?" Her answer reveals whether she's locked into Porto long-term or hedging — and that calibrates everything from the listings shortlist to the price-aggressiveness conversation.
Public records, not private dossiers.
Every source in this protocol is genuinely public — anyone can walk into the registry office and pull the same records (most can be queried online for a small fee). This is structurally different from paid people-search services that aggregate breach data, re-sell scraped social profiles, or stitch together private inferences. We don't use those, and you shouldn't either.
The discipline that protects the relationship: the records inform the agent's strategy, but never appear in any client-facing artefact. No “I see you bought your last property under asking” even though it's technically a public fact. The records are how the agent prepares. The relationship is built in the calls and showings, where the records' influence is invisible to the client and felt only in the calibration of the agent's questions and pace.
Pulling the records is step one.
Letting them inform invisibly is step two.
Lumi is the app that runs this workflow for you. You speak after a showing — Lumi captures the soft signals. You forward an email — Lumi updates the constraints. You open the app at 8am — the brief is already there, ready to feed Claude.
- Voice → structured CRM, automatically
- No forms. No data entry. No copy-paste.
- Free for agents in EU · LatAm · MENA
Lumi · Wednesday
Good morning, Niki.
Two showings · three leads need a nudge.
Showing · Passeig de Gràcia 84
Pipeline
Active
8
Warm
4
Cold
2
Clara Ruiz
Active€1.8M · 3BR
Passeig de Gràcia showing · 11:30
Andreas Moreno
Active€2.4M · 4BR
Send comps by 18:00
Dimitri Schneider
Warm€900K · 2BR
Contract review today
Silent 3d · last 3 days ago
Sarah Mitchell
Cold€1.2M · 3BR
Draft re-engagement
Silent 9d · last 9 days ago
A residential adaptation of the due-diligence dossier pattern from commercial RE — public registries + LLM synthesis to walk into the first meeting already knowing the patterns. Our slice: the 4-sentence transaction-history brief, read in 8 seconds before each first call.
More guides like this on @lumi.estate. Follow if any of this was useful — it's how we know to keep writing.