// marketing brain

Every agent reports to one Brain.

The Marketing Brain sits above Magistry's agents. It reads the signals from your store, decides what matters most, and proposes one plan. Then it earns the right to carry it out, lever by lever, on a trust ramp you can audit.

Install on Shopify

See how autonomy is earned

Live One plan across channels Autonomy earned per lever Applied actions scored win / loss

brain · tick — example output

signals        : 7 pending (1 urgent)
context        : performance · pacing ·
                 competitors · seasonality

proposal       : ads_google · raise budget
                 (within envelope, +18%)
proposal       : seo · rewrite metadata
                 on 4 product pages
proposal       : feed · repair 12
                 disapproved items

autopilot      : 1 applied (reversible)
for review     : 2 queued as cards

// what it does

Orchestration, not another automation.

Point tools optimize their own corner and fight each other for the same budget. The Brain sees the whole store and resolves conflicts before anything ships: at most one open action per subject in each pass.

One plan across channels

The Brain reads performance, budget pacing, competitor prices, feed health, tracking coverage and the season, and proposes one plan. In the last 30 days (measured 17 Oct 2026) it ran 926 decision passes and applied 282 actions in real systems, mostly Google Ads budget moves and shared negative keyword lists, SEO metadata rewrites and Merchant Center feed repairs.

Autonomy is earned, per lever

Each lever, such as a budget move or a negative keyword list, climbs its own trust ramp. At least 6 measured outcomes, a win rate of 70% or more, a recent win and no loss in the last 14 days earn it one autonomy tier. A win rate under 40%, or 3 losses in 14 days without a win, sends it back to your review. Fast down, slow up.

The verdict engine keeps score

Every applied action is scored win, loss or neutral once its full measurement window has passed, against your store's break-even ROAS and adjusted for how the whole store moved in the same window. Confidence is checked against realized win rates, and that check can only add review, never remove it.

A budget envelope it cannot break

Per-domain soft daily spend caps, and a budget move is clamped to ±20% of the current budget. Increases that would breach the envelope stay proposals; decreases and pauses always go through. Irreversible actions are never auto-approved, on any tier.

A strategist and a memory

A weekly strategist pass sets a small set of initiatives, each with a thesis it re-checks against current data and closes when the data no longer supports it. A weekly reflection pass turns measured outcomes into lessons the Brain reads on later decisions.

Urgency preemption

The Brain runs a decision pass every 3 hours. A ROAS collapse, a complaint cluster or a reputation crisis wakes it within minutes instead. A self-healing check re-queues a stalled step of the loop and records what it did.

// the autonomy ladder

From dry-run to autopilot, at your pace.

The Brain is off until you turn it on. You choose the mode per store; the trust ramp decides what each lever is allowed to do inside it.

// 01 · dry-run

The Brain proposes and logs everything, executes nothing. You watch its reasoning on your real data, and every proposal carries the evidence it was built on.

// 02 · approval-only

Proposals queue as cards in the dashboard. You approve item by item; approved moves are applied and scored, and that record is what the trust ramp reads.

// 03 · autopilot

Gated, reversible moves auto-apply inside the budget envelope. Everything else still comes to you, and a lever whose record turns is demoted automatically.

trust_ramp · lever view — example

lever              status      window
─────────────────  ──────────  ───────
budget_set         autopilot   earning
negative_keywords  autopilot   earning
rewrite_metadata   review      unproven
feed_repair        review      demoted

promotion rule : ≥6 measured outcomes,
                 ≥70% win rate, no loss
                 in the last 14 days
demotion rule  : <40% win rate, or 3
                 losses in 14 days
irreversible   : always your call

// works with