Ghost Engine Field Guide
Verified portfolio map · 2026-07-28

Ghost Engine,
mapped to the portfolio.

A detailed operating guide to what Ghost Engine is, how it works, where it fits across local-search projects, and the safety gates between a useful pilot and a burned account.

6 owned properties85 lead-gen projectsAgency client candidates
GHOST ENGINE
Ghost
Hybrid
Speed
6 owned properties
85 lead-gen projects
Client candidates
7 integrations
01 · The engineWhat it is

A local-search behavior operating system

A local-search behavior operating system

Ghost Engine is Brian Costello's Windows-first Node engine for controlled browser behavior around local search. It combines personas, residential proxies, geo-matched fingerprints, CTR sessions, account warmup, rank tracking, geogrids, publishing, scheduling, and reporting behind one API and CLI.

  • Verified local project: version 2.0.0 with 109+ plugins and 704 documented routes.
  • Three modes: ghost for trusted OS input, hybrid for OS input plus browser reads, and speed for lower-risk CDP work.
  • The engine is infrastructure. Accounts, proxies, warmup, realistic pacing, and measurement determine safety.
Detection-test scores are not proof of real-platform success. Pilot, measure, and preserve holdouts.
02 · ArchitectureHow it works

Identity, geography, behavior, and evidence stay coherent

02 · Architecture

Evidence → controlled action → measured result

Every run starts with a real property, query, city, approved account pool, and matching residential proxy. The engine selects an input mode, performs human-paced behavior, records events, updates account health, and feeds the next schedule.

  • Identity: encrypted cookies, deterministic fingerprint, behavioral profile, and account health.
  • Geography: proxy city, timezone, coordinates, language, and UULE must agree.
  • Behavior: search, click, dwell, scroll, navigation, and realistic exit patterns.
  • Evidence: rank movement, geogrid change, calls, conversions, failures, and health.
Desktop ghost mode is foreground and focus-bound. It is not an unlimited parallel worker.
03 · Best fitSix controlled properties

Owned GMBs are the safest first campaigns

03 · Best fit

Evidence → controlled action → measured result

Owned properties are the strongest candidates because you control the brand, site, GMB, tracking, and operational risk. Start with one measured pilot rather than activating the whole portfolio.

  • Moving-services property A
  • HVAC property
  • Cleaning-services property A
  • Moving-services property B
  • Cleaning-services property B
  • Roofing property
Pick the property with the cleanest CallRail or GMB attribution and a stable ranking baseline.
04 · Portfolio scale85 project folders

Run the lead-generation portfolio by family and city pod

04 · Portfolio scale

Evidence → controlled action → measured result

The verified F-drive inventory contains 85 top-level lead-generation projects. Treat them as campaign families with shared governance, not as 85 unrelated automations.

  • Roofing: Aurora Shield, Celestia Crest, Elite Roof Leak, Skyline, All Season, Mark Roofing, and city properties.
  • Electrical: Big State, Cricket Residential, and Excel Electricians across multiple metros.
  • HVAC: HVAC property, HVAC Repair Chief, and city repair properties.
  • Plumbing: Peerless Plumbing and Dallas-area city properties.
  • Restoration: mold remediation and water-damage properties.
A folder proves a project exists. It does not prove the site, GMB, tracking, or authorization is currently live.
05 · Approval gateClient assets

Measure first, automate only after written approval

05 · Approval gate

Evidence → controlled action → measured result

Client projects can use rank tracking, geogrids, account-safe publishing, and scoped CTR tests. Reviews, account creation, social posting, and behavioral campaigns require explicit client and risk approval.

  • Solar client: rank, geogrid, controlled local-search pilot, and CallRail attribution.
  • Multi-location home-service client location campaigns measured separately.
  • Multi-location recovery client tracking and publishing.
  • MDAI local clients: 12 folders spanning local service, legal, medical, dental, therapy, and remodeling.
  • YMYL legal and health categories require stricter claims, review, and identity governance.
Folder presence never equals campaign authorization.
06 · Control surfacesInternal products

Connect through the API instead of copying engine logic

06 · Control surfaces

Evidence → controlled action → measured result

Several internal products are natural status, scheduling, attribution, and proof surfaces. They should call authenticated Ghost Engine routes and consume events rather than becoming competing engine implementations.

  • Product wrapper: confirmed productized wrapper around the browser and automation core.
  • Browser Ops Dashboard: sessions, modes, proxies, and event visibility.
  • Bird's Eye ROI: rank, geogrid, call, and conversion attribution.
  • Competitor Watch: baselines and movement alerts around target queries.
  • LocalForge, Pleper Tools, and Domain Portfolio Dashboard: property setup and campaign inputs.
The canonical source repository is private and intentionally omitted from this public explainer.

Preflight before traffic

Choose one owned property with clean attribution, capture the baseline, approve caps and stop conditions, then run the smallest useful pilot.

Open the operator runbook →