a Sagan program · currently in private beta

Machine Hunter Agent

Your dashboard consolidates HiBid, ProxiBid, BidSpotter, GovDeals, and four other auction sites plus Craigslist into deduplicated rows with extracted specs, filtered by your SOP keywords, so you approve qualified lots into Google Calendar.
before

You open HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist every morning. For each candidate lot, you eyeball photos, decode the serial number to confirm year of manufacture, read model suffixes to confirm phase and cooling, and manually add the auction end-date to your calendar. The same listings appear across sites. Deadlines slip past unnoticed. You've tried offloading this to a remote employee, but the spec-decoding is too granular for them.

after

You open one dashboard every morning. All qualified Taylor, Carpigiani, Electro Freeze, and Stoelting machines from eight sources appear in deduplicated rows with extracted specs. You approve machines one click at a time. Each approval writes the auction end-date into your Google Calendar with the lot URL. You can write rules like 'skip silver machines' or 'only show Electro Freeze after 2015' and the dashboard re-filters in real time.

food service equipment / equipment wholesale / machinery resale / auction aggregation / equipment refurbishment / lead intake / data enrichment / operations / scheduling / HiBid / PCI Auction / ProxiBid /  food service equipment / equipment wholesale / machinery resale / auction aggregation / equipment refurbishment / lead intake / data enrichment / operations / scheduling / HiBid / PCI Auction / ProxiBid / 
the problem

Stop hunting Taylor and Carpigiani machines across eight auction sites every morning.

Equipment refurbishers waste hours daily opening HiBid, ProxiBid, BidSpotter, GovDeals, and four other auction tabs plus Craigslist to eyeball serial plates, decode model suffixes, and manually add end-dates to a calendar.

01
Daily manual sweep across 8 sources

HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist alerts.

02
Duplicate listings across platforms

Same lot appears on HiBid and the auctioneer's own site. No deduplication.

03
Spec decoding by hand

Reading serial plates, confirming year of manufacture, checking phase and cooling type from model suffixes.

the math, if you want to look

One dashboard replaces your daily hunt across eight auction sites.

proof 01
Consolidated dashboard

All qualified listings from HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist in one view.

proof 02
Automatic deduplication

Same lot appearing across multiple sources is shown once, with source links for comparison.

proof 03
Extracted specs

Brand, model, serial number, year of manufacture, phase, and cooling type parsed from photos and listing text.

proof 04
Your rules, your filter

Write plain-English rules like 'skip silver Taylor 3-series' or 'only Carpigiani LB-series after 2015.' The agent re-filters in real time.

proof 05
Calendar integration

Approve a listing, the auction end-date drops into your Google Calendar with the lot URL.

Machine Hunter scrapes all eight sources, deduplicates listings by photo hash and title, extracts brand, model, serial number, year, electric phase, and cooling type from data-plate images and description text, filters against your brand and spec rules, and shows only qualified Taylor, Carpigiani, Electro Freeze, and Stoelting lots in one dashboard. One click writes the auction close-date into your calendar with the lot URL.

how it works

How Machine Hunter works

The agent monitors eight auction platforms and Craigslist, extracts specs from every listing, filters by your rules, and surfaces only qualified machines in one dashboard.

step 01
Scrape all eight sources every 4 hours

HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist saved-search email alerts feed into the system.

step 02
Deduplicate across sources

Same lot appearing on HiBid and the auctioneer's own site is recognized as one listing. Dedup uses photo perceptual hashing, title fuzzy matching, price, and location.

step 03
Extract specs from photos and text

Vision AI reads data-plate photos to extract serial number, model, year, phase, and cooling type. Listing description text is parsed for the same fields. Year of manufacture is decoded using brand-specific serial-number rules (Taylor's first character and second digit encode decade and year-within-decade; other brands use description text).

step 04
Apply your rules

You write plain-English rules like 'skip silver machines' or 'only show Electro Freeze after 2015.' A lightweight AI model evaluates each rule against the extracted fields and listing text. Filtered-out listings show a trace so you can see why they were hidden.

step 05
Show qualified listings in one dashboard

Only machines matching your brands and rules appear. Each row shows the lot title, extracted specs, photos, location, auction end-date, source link, and a yes/no button.

step 06
Approve to calendar, reject to refine rules

Click yes, the auction end-date and lot URL drop into your Google Calendar. Click no, you can update a rule so similar listings don't show up again.

ai agent · estimator console inputs transform outputs public preview
inputs
HiBid auction listings +

Lot title, description, photos, high bid, end-date, auctioneer name.

ProxiBid auction listings +

Lot title, description, photos, location, end-date.

BidSpotter auction listings +

Lot title, description, photos, location, end-date.

GovDeals auction listings +

Lot title, description, photos, location, end-date.

TagEx / Restaurantequipment.bid listings +

Lot title, description, photos, location, end-date.

AuctionFactory listings +

Lot title, description, photos, location, end-date.

Bay Area Auction Services listings +

Lot title, description, photos, location, end-date.

Craigslist saved-search email alerts +

Listing title, description, photos, seller contact, posting date.

Your brand and spec rules +

Plain-English rules you write: 'skip silver Taylor 3-series', 'only Carpigiani LB-series after 2015', 'auto-reject China or Vietnam'.

transformation
Scrape all eight sources every 4 hours +

HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services via web scraping. Craigslist via inbound email parsing of saved-search alerts.

Deduplicate across sources +

Same lot appearing on multiple platforms is recognized as one listing using photo perceptual hashing, title fuzzy matching, price, and location.

Extract specs from photos and text +

Vision AI reads data-plate photos for serial number, model, year, phase, and cooling type. Listing description text is parsed for the same fields. Year of manufacture is decoded using brand-specific serial-number rules (Taylor: first character and second digit encode decade and year-within-decade; other brands: extract from description text).

Filter by your rules +

A lightweight AI model evaluates each rule you write against the extracted fields and listing text. Listings matching your brand list and passing all rules are shown; filtered-out listings display a trace explaining why.

Rank and display in dashboard +

Qualified listings appear in deduplicated rows with extracted specs, photos, location, auction end-date, source link, and yes/no buttons.

outputs
One consolidated dashboard +

All qualified Taylor, Carpigiani, Electro Freeze, Stoelting, Emery Thompson, Cattabriga, Technogel, and C. Nelson machines from eight sources in deduplicated rows.

Extracted specs per listing +

Brand, model, serial number, year of manufacture, electric phase (1PH or 3PH), cooling type (air or water).

Approval/rejection workflow +

Yes/no button per listing. Yes creates a calendar event. No prompts you to update a rule.

Google Calendar events +

One event per approved auction with the lot title, end-date, and lot URL in the description.

Rule refinement interface +

Plain-text rule editor where you write and update rules like 'skip silver machines' or 'only show Electro Freeze after 2015.' Dashboard re-filters in real time.

draft ready for estimator review _
tech used
HiBid, PCI Auction, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services auction platformsCraigslist saved-search email alertsZenRows web scraping with premium proxy and JavaScript renderingApify actors for specialized auction sourcesOpenRouter vision and lightweight AI models for spec extraction and rule interpretationGoogle Calendar API for auction deadline schedulingSQLite for listing deduplication and rule persistence
tool alternatives
Apify actors as fallback for HiBid and GovDeals scrapingEmail-based calendar integration (.ics files) as alternative to Google Calendar API
honest qualification

Is this for you?

built for you if
  • + Equipment refurbishers and resellers - Companies sourcing used soft-serve, batch-freezer, gelato, milkshake, and pasteurizer machines from multiple auction platforms and Craigslist.
  • + Owners and procurement teams - Decision-makers who personally hunt inventory or oversee sourcing teams and need to consolidate eight sources into one dashboard.
  • + Businesses with brand and spec requirements - Companies that buy only specific brands (Taylor, Carpigiani, Electro Freeze, etc.) and need to filter by year, phase, and cooling type.
  • + Operations with calendar-driven workflows - Teams that track auction end-dates and need to route approved lots into a shared calendar for bid management.
not for you if
  • - Businesses that source from a single auction platform - If you only use HiBid or one other source, the consolidation benefit is minimal. The agent is built for multi-source hunting.
  • - Buyers with no spec requirements - If you buy any machine regardless of brand, year, or phase, the filtering and rule layer add no value.
  • - Craigslist-only sourcing - The agent handles Craigslist via saved-search email alerts, which carry a 1–6 hour lag. If you need real-time Craigslist monitoring, this is not the right tool.
  • - Facebook Marketplace sourcing (v1) - Facebook Marketplace is deferred to v2. v1 covers HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist only.
pricing

Pricing

to build

Machine Hunter is a scoped custom build. Pricing covers the initial build and integration with your eight sources and Google Calendar. Ongoing costs are usage-based: web scraping, vision AI for data-plate extraction, and lightweight AI for rule filtering.

then
  • Initial build includes HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, Craigslist email parsing, and Google Calendar integration.
  • You can test the full workflow with sample listings before connecting your live auction accounts or calendar.
  • Usage-based costs scale with refresh frequency (default every 4 hours) and listing volume. Typical monthly cost depends on your refresh cadence and the number of listings processed.
  • Facebook Marketplace is deferred to v2 and priced separately.
  • Custom rules and brand-specific serial-number decoders are included in the initial build.
FAQ
How does Machine Hunter handle the same auction lot appearing across multiple platforms?

Machine Hunter deduplicates listings across all eight sources using photo perceptual hashing, title fuzzy matching, price, and location. When the same lot appears on HiBid and the auctioneer's own site, it shows as a single row in your dashboard with links to both sources so you can compare.

What information does Machine Hunter extract from each listing?

The agent extracts brand, model, serial number, year of manufacture, electric phase (1PH or 3PH), and cooling type (air or water) from both listing descriptions and data-plate photos. Year of manufacture is decoded using brand-specific rules. For example, Taylor serial numbers encode the decade and year in the first two characters.

Can I write my own filtering rules, or are the filters fixed?

You write and edit the rules yourself in plain English. Examples: 'skip silver Taylor 3-series machines' or 'only show Electro Freeze after 2015.' The agent re-evaluates all listings against your rules in real time, and you can see why each listing was filtered out.

What happens when I approve a listing?

Clicking yes on a listing writes the auction end-date and lot URL into your Google Calendar as a single event. You can then manage your bids from your calendar. Clicking no drops the listing and optionally prompts you to update a rule so similar listings don't show up again.

Which auction platforms does Machine Hunter monitor?

Machine Hunter monitors HiBid, PCI Auction, ProxiBid, BidSpotter, GovDeals, TagEx/Restaurantequipment.bid, AuctionFactory, Bay Area Auction Services, and Craigslist saved-search alerts. Facebook Marketplace is planned for a future release.

How often does Machine Hunter check for new listings?

The agent refreshes all eight sources every 4 hours by default, pulling new listings and re-filtering them against your rules. This cadence keeps you ahead of fast-closing auctions without overwhelming your dashboard with stale data.

What if a listing is missing a serial number or data plate photo?

Machine Hunter flags listings with incomplete specs as 'needs more info' so you can decide whether to investigate further. You can still approve these listings to your calendar if you want to follow up with the seller directly.

Is Machine Hunter built for single-source monitoring, or does it require multiple auction platforms?

Machine Hunter is built for multi-source hunting. If you only use one auction platform like HiBid, the consolidation and deduplication benefits are minimal. The agent delivers the most value when you're monitoring multiple sources and need to avoid duplicate bids and missed deadlines.

next step

Stop hunting machines across eight auction sites every morning.

Machine Hunter consolidates HiBid, ProxiBid, BidSpotter, GovDeals, TagEx, AuctionFactory, Bay Area Auction Services, and Craigslist into one dashboard with extracted specs, filtered by your rules, and routed to your calendar in one click.