Trust & security

Built on a trust-first architecture.

FeedGraph is designed around the principle that merchant data belongs to the merchant. This page documents the controls, cryptographic primitives, and governance model that back that commitment.

16
documented controls
SHA-256
hashed API keys
Hash-chained
audit log
HMAC
signed webhooks
Connection security

Authentication & transport controls

How FeedGraph authenticates to your platforms, scopes what it can touch, and ensures every event and webhook can be verified end-to-end.

01
Scoped Shopify connection

FeedGraph never asks for your Shopify password. You connect either through Shopify OAuth, or with a dedicated app’s client credentials that you create and can revoke — both scoped to the minimum permissions FeedGraph needs: read and write products and inventory, nothing else. Stored credentials are encrypted at rest.

02
Transparent permissions

Every permission FeedGraph requests is shown to you during setup. Nothing is requested in the background or added after the fact without your explicit approval.

03
Read-only ad access by default

FeedGraph reads performance data from Google and Meta and cannot write to either. Writing is governed by a per-workspace Write Actions switch that is off until you turn it on — and turning it on still does not let anything through without per-change approval.

04
Scoped API keys

API access uses SHA-256 hashed keys with defined permission scopes and optional expiry. Keys can be rotated or revoked at any time without disrupting other integrations.

05
HMAC-signed webhooks

All outbound webhook deliveries are signed with HMAC-SHA256 so your receiving systems can verify authenticity and reject spoofed or replayed requests.

06
One door out, and it fails closed

Every outbound write to an ad platform is forced through a single enforcement point in the codebase. If the approval plumbing is missing the request fails rather than proceeding — there is no code path that reaches Google or Meta around it.

Data controls

Your data stays yours

Ownership, portability, and the limits of what FeedGraph does with your catalog data — including the right to walk away at any time.

07
You own your data

Your catalog data belongs to you. FeedGraph does not claim any rights over merchant product data and will never use it to train shared models or sell it to third parties.

08
No data sales

We never sell merchant data. Data is shared only with the infrastructure and AI subprocessors required to deliver the specific features you have activated.

09
Export your data at any time

Your catalog and enrichment data can be exported from FeedGraph at any time in standard formats. You are never locked in.

10
Revoke access at any time

You can disconnect FeedGraph from your Shopify admin at any time — no FeedGraph login required. Access is revoked the moment you remove the integration from Shopify.

AI governance

You decide what changes.

Nothing reaches your ad accounts without a person approving that specific change. AI proposals are previewed with their credit cost, run on an approved sample first, and every applied change is logged and revertible.

11
Human review before activation

Bulk AI enrichment runs on a sample you approve before it touches the rest of the catalogue, and quotes its credit cost before it starts. Bid changes show the current value against the proposed one and are capped at a 50% move in a single step. The AI proposes; you decide what gets applied.

12
Nothing reaches a platform without approval

No change reaches Google or Meta without someone approving that specific change, staged with a plain-English summary of exactly what will happen. Catalogue edits inside FeedGraph apply when you make them — and every one is logged and revertible field by field.

13
Every action attributed

Know exactly who made which change, when, and why — across your entire team. AI-originated changes are labelled as such and distinct from human edits.

Audit & change management

Complete, tamper-evident history

Every change — whether by a team member or an AI recommendation — is logged with attribution, timestamped, reversible, and hash-chained so tampering is detectable.

14
Full, tamper-evident change history

Every modification is stored with a timestamp, reason, and the value before and after. Entries are hash-chained — each hashed together with the one before it, so altering or removing any entry breaks every hash after it, and one click verifies the whole chain. The log exports to CSV with the hashes included, so you can verify it independently of us.

15
Reversible actions

Any enrichment or feed change can be reviewed and reverted to the previous state. Reversions are themselves logged — there is no way to hide what happened.

16
Workspace-level permissions

Control what each team member can see and edit at the workspace level. Role boundaries are enforced server-side — client-side UI restrictions are not your only guard.

Responsible disclosure

Found a vulnerability?

If you discover a security issue, please contact us directly before public disclosure. We aim to acknowledge reports within 48 hours and will work with you on a coordinated disclosure timeline.

Report a vulnerability →
Compliance questions

Need more detail?

For compliance reviews, vendor security questionnaires, or architecture deep-dives, reach out directly. We can share additional documentation under NDA for enterprise evaluations.

Contact the team →