CipherGuard Labs / Built by us
Live in production
Product — Reputation Operations

RepuPilot keeps every
storefront's reputation
on autopilot.

A reputation operations platform for multi-location retail. It watches every listing, drafts every reply, and reports on every location — so a five-person team can run reputation for fifty storefronts the way one person runs it for one.

35+
LOCATIONS UNDER MANAGEMENT
24/7
REVIEW & MENTION SCANNING
10/MIN
LISTING EDITS, RATE-MATCHED TO GOOGLE
01 — THE MECHANISM

How a review becomes a reply

Every review, mention, and listing change moves through the same flight path — no location is ever handled by hand unless it needs to be.

WAYPOINT 01
Scan
Pulls every review from every location directly from Google, and tracks brand mentions across dozens of Reddit communities in parallel.
WAYPOINT 02
Draft
An AI model reads the tone and substance of each review and writes a reply in the brand's voice — never a generic template.
WAYPOINT 03
Sync
Approved replies and listing changes — hours, closures, descriptions — post straight back to Google, one location at a time, in order.
WAYPOINT 04
Learn
Sentiment, response time, and staff mentions roll up into a live scoreboard, so every location can be compared on the same terms.
02 — WHAT IT DOES

Built for the operator, not the single storefront

Review aggregation
Every location's reviews land in one place, refreshed on a schedule — no manager checking each listing by hand.
AI-drafted replies
Every review gets a considered response, drafted in the brand's voice and ready for a one-tap approval.
Auto-reply for the routine cases
Straightforward five-star reviews get answered on their own, on a natural delay — the team's time goes to the reviews that need it.
Reddit & mention monitoring
Brand mentions across dozens of communities are caught as they happen, not weeks later.
Listing management at scale
Hours, descriptions, and temporary closures update across any number of locations from a single screen.
Reporting built for many locations
Sentiment, response rate, and performance by location — the view a multi-location operator actually needs.
03 — UNDER THE HOOD

The mechanism, in engineering terms

No shortcuts on the parts that matter: rate limits, review coverage, and who's allowed to touch a listing.

Backend
Node.js service on Railway, backed by PostgreSQL for every location, review, and reply record.
Job queue
BullMQ handles scheduled review syncs, delayed auto-replies, and scheduled listing changes — so nothing depends on a person clicking a button on time.
Review coverage
Paginated, per-location retrieval against the Google Business Profile API — built to pull a location's full review history, not just the most recent batch.
Listing control
Writes go through the Business Information API, throttled to Google's own edit limits per profile — so bulk changes never trigger a lockout.
Social listening
RSS-based scanning across dozens of Reddit communities, tuned for retry and timeout handling so a single dead feed never stalls the rest.
AI layer
Claude drafts every reply and reads reviews for sentiment and staff mentions — tuned for cost efficiency at review volume.
Access model
Team members sign in with their own Google or Microsoft account. Google's review and listing access is connected once by an admin and shared safely across the whole team — nobody re-authorizes Google to use the platform.