// works with
Every one of these is both a data source and a place Magistry writes back to — no new storefront, no migration, no glue code.
Shopify, Google Ads, Merchant Center, Google Analytics 4, Meta, Instagram, TikTok, Pinterest, Microsoft Bing, Email · IMAP/SMTP
// command
Not read-only integrations with a suggestion box on top. Magistry holds working control of every commercial lever a store has — and uses it.
Every price, every variant, every lifecycle state — moved on your real margins, not a rule of thumb.
Budgets, bids, audiences and whole campaigns, across Google and Meta — paced on profit.
Reads, answers, refunds and resolves — in your voice, grounded in your own policies.
Refunds, gift cards, cancellations — every euro move weighed against what it protects.
What Google, Meta and TikTok are shown of your catalog — per market, kept eligible.
Copy, content, promotions, experiments — down to the theme code itself.
Drafted, scheduled, published and answered across six networks, in one brand voice.
Segments built from your own data and synced to wherever the ad money is spent.
If it's part of running a store, Magistry can run it. And everything it runs, it can undo.
// the daily grind
Not strategy — upkeep. The recurring checks and fixes that keep revenue from leaking. Magistry takes exactly this work, job by job.
Open every campaign, read the curves, pause what's bleeding — and hope you caught it before lunch, not after.
// with Magistry
Budgets and bids paced against profit, anomalies auto-paused, search terms returned with a keep-or-negate verdict.
campaign_agentDisapprovals, policy flags and stale feeds across five channels — each one quietly throttling reach until someone notices.
// with Magistry
Merchant Center's verdict pulled back per market, attribute gaps scored, fixes queued in the order they cost you impressions.
feed_agentWhere's-my-order, sizing complaints, refund requests. Every reply needs the order, the policy, and your brand voice.
// with Magistry
Threads classified and matched to their order, replies grounded in your policies and scored against your voice before they're offered.
cs_agentDraft the dead stock, reprice the movers, vault the seasonal lines, and catch the products cannibalising each other.
// with Magistry
Every SKU in one lifecycle state, moved on your margins, with the reverse operation stored before anything is written.
catalog_agentThe competitor launch, the demand spike, the new supplier — usually spotted a month after it mattered.
// with Magistry
News, social and search demand scanned continuously, signals credibility-scored, and research lanes surfacing your next product.
market_expertThe loop runs overnight, on weekends, across every store you own. You review decisions, not dashboards.
// the stack
This is the stack a serious Shopify store ends up paying for. Every tool solves one job well and then keeps its data. The helpdesk never sees the margin. The feed tool never sees the ad spend. The attribution tool never sees why a customer wrote in.
swipe to read the full ledger →
| The job | What you pay for it now | List price / mo | In Magistry |
|---|---|---|---|
| Customer service | Gorgias · Zendesk · Re:amazeper ticket tier, plus ~$0.90 per AI resolution | $60 – $900 | CS Specialist + inbox |
| Attribution & analytics | Triple Whale · Northbeamscales with your GMV | $219 – $2,000+ | Tracking & Attribution |
| Server-side tracking | wetracked.io · Elevarper monthly order band | $49 – $249 | Server-side events + CAPI |
| Shopping feed management | DataFeedWatch · Channable · Feedonomicsper SKU count and channel | $64 – $239 | Feed Syndication |
| Profit analytics | TrueProfit · BeProfit · Lifetimelyper order volume | $25 – $100 | Pulse profit breakdown |
| Chargeback recovery | Chargeflow · Signifydsuccess fee on every dispute you win | 25% of recovered | Dispute Shield |
| Product & competitor research | Minea · PiPiADS · Dropship.ioper seat and credit pack | $39 – $159 | Researcher + Market Expert |
| Reviews & reputation | Okendo · Loox · Judge.meper monthly order band | $15 – $299 | Reputation |
| Social scheduling | Later · Buffer · Hootsuiteper channel and seat | $25 – $99 | Social Autopilot |
| Storefront experiments | Intelligems · Visuallyper order volume | $15 – $80 | Theme Studio experiments |
| Ten vendors | Ten logins, ten billing dates, ten support queues — and no shared record of what anyone changed. | $500 – $4,000+ | One platform |
Public list prices, checked July 2026. Nearly all of these bill on order volume, GMV, tickets or seats, so a growing store's real invoice sits at the top of each range — and the chargeback fee is a cut of money you recovered.
The bill is the small problem. The real one is that no tool in that list can answer “is this product still worth the ad spend, given the refunds it generates and the margin after shipping?” — because no tool in that list has all three numbers. Magistry does, which is the only reason it can decide instead of report.
// the product
Eight screens from the live product — real UI, demo store data. No mockups, no marketing renders: these are the surfaces your team would work in on day one.
Pulse leads with profit-true numbers — net margin after COGS, refunds, ad spend and operating costs — and then hands you Otto. Ask in plain language; Otto reads the live data, answers with citations, and proposes writes that wait for your approval.
// the marketing brain
Most AI tools hand you an answer and hide the reasoning. The Brain renders its whole working memory — what it knows, what it's trying, what it has won and lost, and what it wants to do next — as one graph you can click into.
// working memory
Actions are sized by expected impact and fade with age. Anything waiting on you wears a dashed approval ring, so the graph itself tells you where to look.
// data senses
Each data stream — orders, ads, costs, catalog, feed, service, tracking, reviews, competitors — orbits the core with a live richness score, and says what it would unlock if you fed it.
// track record
Domains and initiatives carry win/loss arcs from the full history. The Brain leans into the levers that have worked in your store and stays cautious on the ones that haven't.
// one budget
Signals from every agent land in a single decision pass inside one budget envelope — so ads, pricing, catalog and feed moves are weighed against each other, not in isolation.
// the engine
So Magistry was built to see all of it. Every stream below is weighted by how much it actually changes a decision, and the more you connect, the less it has to guess — that is the entire mechanism by which a store gets more profitable.
// what it reads
Weights are the real ones the platform scores with. Connect a stream and the Brain tells you what it just unlocked.
// what it does with it
Digest
Every signal above lands on one bus with a severity, a source and a subject — not in a per-tool dashboard.
Plan
One decision pass across every domain inside one budget envelope, so ads, pricing, catalog and feed moves compete against each other instead of running blind.
Judge
Each move is checked against policy, the tier of evidence behind it, and the store's own calibrated break-even before it is allowed to write.
Execute
The write goes out with its reversal operation already stored, and the row lands in the append-only log.
Score
The outcome is measured and marked win, loss or neutral — and the Brain's confidence in that lever moves accordingly.
Run every cycle, on a schedule, whether or not anyone is watching.
// what moves
Every one of these is a reversible row. See the log →
// otto
The colleague who read everything before the meeting. Ask a question, get an answer with the rows cited. Ask for a change, get it executed — against your live store, inside your rules.
A real exchange — tool calls, a cited answer, and the approval card that gates every write.
// difference
The distance between a suggestion and an outcome is execution. Otto closes it — carrying the full context of your store, and leaving a record of every move it made.
Otto writes to Shopify, Google Ads, Meta, TikTok, and your inbox — directly. Not a chat wrapper. Not a CSV exporter. Every action lands in the system that runs your store.
Before every move, Otto reads your config, your decisions, your evidence, and your guardrails. It picks up exactly where last night's run left off — no warm-up, no re-prompting.
Approve a decision row once, and Otto runs the same pattern — audited, gated, reversible — on a schedule. No glue code, no Zaps, no copy-paste from a chat transcript.
// how it works
No agent gets write access on day one. Each starts read-only, builds a track record you can inspect, and widens exactly as far as its results justify.
Install from the Shopify admin in minutes. Magistry syncs your catalog, orders, costs, and ad accounts — secrets vault-encrypted from the first byte.
OAuth install · no code · no credit card
Every agent starts in dry-run: decisions land in the log with full reasoning and evidence, but nothing is written to your store until you promote.
decision_log fills · zero writes
Start in approval mode and accept decisions one by one. When the rows keep matching your gut, widen autonomy per agent — and roll back any row, any time.
per-agent modes · ↶ rollback · kill switch
Free plan · No credit card · Starts in dry-run
// evidence
A number without its derivation is a guess. Every row names the signals it read, the window it read them over, the confidence it landed on — and the reversal it stored before writing anything.
Lifecycle transitions across the catalog, validated against ALLOWED_TRANSITIONS, applied in batches with reversal pre-computed.
Budget pacing and bid moves cited by anomaly windows, judged against policy, gated by tier-of-evidence before any euro moves.
Operator-owned kill switch, per-action rate limits, advisory locks, vault-encrypted secrets, and a monthly LLM budget cap.
Decision-row acceptance, dry-run-to-live ramp curves, judge agreement rates — surfaced as the metrics that tell you when to widen autonomy.
decision_log, ad_decision_log, cs_thread_evaluations — every transition is a row with reason, evidence, citation, and the reversal op pre-stored.
// enterprise readiness
A platform that reads this much of your business gets asked hard questions before it is allowed near it. Here are the answers, in the order security teams ask for them.
Primary data stores run on EU-region infrastructure. Encrypted at rest with AES-256, TLS 1.3 in transit — including service to service.
A full sub-processor list, documented retention, and export or erasure on request. Card data never touches Magistry — Shopify handles payments.
Your catalog, orders and customer conversations are never used to train models. Inference runs under no-training terms with every provider.
RBAC on every route and every write, per-tenant isolation enforced in the database, and advisory locks so two operators can't collide.
Every channel token lives encrypted in Supabase Vault, scoped per store. Nothing is stored in plaintext, and nothing is shared across tenants.
Every external write is a row with its reason, its evidence, its author and its reversal op. Rows are never edited and never deleted.
And the one we can't claim yet: SOC 2.
We're early-stage and not certified today. We build to the same controls — least privilege, encryption, audit logging — and we'll walk a security team through the current posture line by line.
// the platform
Growth, operations, and defense on one auditable spine — coordinated by the Marketing Brain, not stitched from point tools.
// grow revenue
Budgets and bids paced on profit, not on the ROAS the ad platform grades itself. One product feed syndicated to five channels. And underneath it, your own server-side tracking — so every verdict rests on numbers nobody else marked.
// run lean
Every SKU moved through its lifecycle on your real margins. Every customer email answered in your voice, grounded in your policies, refunds gated by expected value. And the storefront itself tested, edited and rolled back under approval.
// protect & discover
Chargebacks fought with assembled evidence. Reviews and mentions classified and answered before they harden into reputation. And a research engine reading ad libraries, marketplaces and search demand for the product your competitor hasn't found yet.
// the fleet
Here they are colleagues. Each runs its own plan-judge-execute loop against your live systems, and all of them answer to the Marketing Brain — so the store gets one plan, not four opinions.
// Runs the lifecycle of every SKU.
Drafts underperformers, archives dead stock, runs margin-aware discounts, and publishes winners — every product in exactly one lifecycle state with a reversible row attached.
// writes to
// mandate
Full authority over every SKU’s lifecycle — draft to publish to vault and back.
// Operates your paid acquisition.
Four specialist skills — Bid & Pacing, Budget Allocator, Lifecycle Manager, Compliance & Guardrails — wired into a planner-judge-executor loop per campaign, executing on Google Ads.
// writes to
// mandate
Runs paid acquisition end to end — budget, bid, audience and creative.
// Handles inbound end-to-end.
Order Detective + Policy RAG + Reply Drafter + Reply Judge. Brand-voice embeddings hold tone; language-gated send; proactive outbound on in-flight issues before the customer asks.
// writes to
// mandate
Owns a conversation from first read to closed ticket — money moves included.
// Watches the market for you.
Scans news, social trends, and search demand across research lanes; publishes credibility-scored signals the other agents subscribe to — demand spikes, niche refreshes, geo-expansion windows.
// writes to
// mandate
Watches everything your competitors would rather you missed.
The real gallery — the four above lead a fleet of fifteen, every one with its own schedule, scorecard and autonomy mode.
// beyond the dashboard
Drafts, schedules and publishes across six networks in your brand voice, replies to comments and DMs, and pulls what actually landed back into the same decision log as everything else.
See Social AutopilotAn agentic editor working on a sandbox copy of your live theme. Every change lands as a diff approval card, deploys only on your approve — with snapshot-first rollback and storefront experiments to promote the winner.
See Theme StudioDrive your whole store from your own AI. A remote MCP server for Claude, Cursor, or any MCP client, plus REST tool endpoints with OpenAPI — OAuth 2.1 or scoped API keys, writes gated by approval cards.
See the API// integrations
Per-store OAuth for every channel, tokens in an encrypted vault, and one gate in front of every outbound write. No CSV uploads, no Zapier duct tape.
// magistry standard
Under the Shopify wiring, Magistry reasons about customers, an offering, costs, acquisition and service — which is every business, not just retail. Magistry Standard runs the same Brain, the same decision log and the same guardrails against a canonical business model instead of a storefront.
You expose one read-only endpoint against a contract we own and version; Magistry pulls, validates and starts reasoning. Paid acquisition, tracking, customer service and the autonomy ladder work exactly as they do for a store.
Onboarded hands-on, not self-serve — the connector is built once, with us.
// what changes
// partner program
Agencies earn recurring revenue-share on every client store — 20% rising to 30%, lifetime, with deal registration and co-marketing funds. Merchants can refer a friend and both sides get AI-credits.
// plans
All paid tiers unlock the whole platform — you grow by AI-credits, nothing else. And there's no trial to expire: the Free plan is your trial.
Free plan · No credit card · Starts in dry-run
// faq
The questions buyers actually ask, answered without hedging. Still missing one? Book a 20-min call.
// magistry
// audit plane
// magistry.io
Connect in minutes and the Brain starts reading immediately — orders, margins, ad spend, feeds, sessions, conversations. The richer it gets, the sharper every decision it makes for you.
Free plan · No credit card · Starts in dry-run
We use cookies to run this site, to see which pages help, and to measure which ads bring people here. You choose per category; everything except the strictly necessary is off until you say otherwise. Cookie policy · Privacy
