Technology
Reads your catalog. Does the math.Keeps to your rules.
Two engines and a set of hard guardrails, built on the official Amazon Ads API and SP-API.
AdPraxis agent
Architecture
Four layers, one closed loop
Every run moves from raw data to a recorded decision. No layer is skipped.
- L4Execute
- L3Decide
- L2Understand
- L1Data
Engine 01
Smart engine
From a product list to a launch-ready account, planned against your own history.
- 01
Semantic keyword mining
Reads titles, bullets and images, then clusters candidates into traffic lanes.
- 02
Budget-aware structure
Picks the finest structure where every campaign can still afford an order.
- 03
Inventory gate
Out-of-stock or stale inventory is blocked before anything is built.
- 04
Verified launch
Created paused, switched on only when every layer is confirmed on Amazon.
- Keywords mined per product
- 146
- Campaigns per submission, max
- 30
Engine 02
Bid decision engine
Every keyword and auto target, every run, judged on more than one bad week.
- 01
Multi-window evidence
7, 14 and 30-day ACOS, ROAS, conversion and CPC, read together.
- 02
Diagnose before acting
High clicks and low conversion read as a listing issue, not a bid problem.
- 03
Risk grading
Every change is scored by size: 5% and 15% split low, medium and high.
- 04
Server re-validation
Each change is re-checked against your shop’s live data. One failure sends the whole proposal back.
- Targets reviewed per run
- 2,400
- Clicks in 7 days before any change
- ≥ 5
Safeguards
Hard limits, built into the engine
- Dry runDefault mode for every new shop
- $0.10–$10Bid band on every change
- 500Actions per proposal, max
- 48 hBefore an unapproved proposal expires
- 1 clickRollback that never overwrites manual edits
- Per shopData isolation, revocable access
Ready to let your ads run by your rules?
A 30-minute walkthrough on an account like yours. No store connection needed.