DealSiftDeal search corpus
Sign in
Every broker site, scraped daily, in one corpus

Search businesses for sale the way you search Google.

Broker-site filters and saved alerts match exact words. DealSift reads every listing across every broker site and finds the businesses that fit what you describe.

New listings added daily · Describe the business you want in plain language
The matches say “turf” & “sod” —
not one listing says “grass”
Use from Claude or ChatGPTDealSift MCP
Human
Are there any artificial grass companies for sale?
Agent
search_deals(“artificial grass companies”) · 14 matches
14 listings fit, across BizBuySell, BizQuest, and businessbroker.net — the brokers write “turf” and “sod”, so keyword alerts miss them. The standout: Verdaline Turf Partners — commercial installs, recurring maintenance contracts, $1.4M EBITDA, on the market 26 days.
Human
Only Arizona, over $1M EBITDA.
Agent
search_deals(…, state: AZ, min_ebitda: $1M) · 2 matches
Two fit: Solquist Turf Co. (Phoenix, $1.2M EBITDA, BizBuySell) and Aravelle Sod & Turf (Tucson, $1.1M EBITDA, BizQuest). The second took a price cut last month — want the history?
Deal names illustrative · Sources real
The alert-email problem

Hundreds of broker alert emails. Which deals you actually see is random.

Anyone sourcing seriously is on every broker’s list, and the deals that get looked at are whichever ones happened to surface in the inbox that day. DealSift makes the search methodical: every site swept daily, every listing scored against your thesis, and seeing a deal no longer depends on which email you opened.

The sites swept daily · 18 sources
bizbuysellbizquestbusinessbroker.netbusinessesforsalecalder capitaldealstreamempire flippersflippaglobalbxlatonasmurphy businessquiet lightsunbeltsynergy business brokersvr business brokerswebsite closerswebsite propertieswoodbridge
CoverageOne record per deal
bizbuysellbizquestbusinessbroker.netsunbeltdealstreamglobalbxSAME DEAL, TWO SITESONE DEAL RECORD“Project Turf”full listing text2 sources · dedupedSCRAPED DAILY

One search covers every broker site you'd check by hand.

The listing sites you sweep every week, scraped daily and deduplicated into one deal record. The anonymous teaser titled “Project Turf” still surfaces, because the match runs on the listing’s full text.

The mechanismHow matching works
HOME SERVICESLOGISTICSECOMMERCEMANUFACTURINGYOUR THESISNEAREST DEALS2,048-DIM EMBEDDING SPACE · 2D PROJECTIONEBITDA / PRICE / GEO = HARD FILTERS

Your thesis becomes a point in vector space.

Listings and your thesis are embedded into the same 2,048-dimension space; the deals nearest your thesis surface first. EBITDA bands, asking price, and geography run as hard filters, never as similarity.

MCPClaude · ChatGPT · your agent
MCP session
$ claude mcp add dealsift <endpoint>
connected · 5 tools

> get_thesis()
  sectors: home services · EBITDA $1–3M · southeast
> search_deals(“HVAC roll-up, recurring revenue”)
  N matches · scored against thesis
> new_deals_since(“last week”)
  new listings · ranked · ready to review

Your AI searches it for you.

Connect any MCP client with one token and it works the corpus directly: reads your saved thesis, runs the narrow searches you never get around to, checks what arrived since last week, pulls a deal’s price history before you write the broker.

ThesisScored on arrival
FIRST SEENDAY 0PRICE CUT−12%RELISTEDNEW BROKERFLAGGEDTHESIS MATCH184 DAYS ON MARKET · HISTORY KEPT PER DEAL

Write the thesis once. Every new listing is scored against it.

Describe what you buy — sectors, size range, geography, what you avoid — and matching listings are flagged as they appear. Each deal carries its history: first seen, price cuts, relistings, real time on market.