01 How it works

It reads across every tool you run, not just one corner.

Every point tool sees its own data and nothing else. Cintrel is the layer that reads across the whole stack you already run, synthesizes it on a schedule, and surfaces the one decision worth making today, with the draft already written. You approve before anything ships.

02 The pipeline

Four quiet steps while you sleep.

Four stages run on a schedule, not on demand. You do not assemble the picture. You arrive to a picture already assembled and a decision already framed.

1
Connect

Cintrel reads by category, not by vendor, across the nine kinds of system a consumer brand already runs. Whatever helpdesk, email tool, or analytics platform you use, it plugs into the same slot. Read-mostly by design: it pulls the data, it does not become another place to maintain. Access is scoped, and every read is logged.

CommerceEmail & SMSHelpdeskWeb analyticsPaid searchPaid socialOrganic searchProduct feedFinance

Live today: Shopify, Klaviyo, Gorgias, GA4, Google Ads, Meta Ads, Search Console, Merchant Center, QuickBooks. Other tools in each category wire in the same way.

2
Synthesize

Each cycle builds a fresh snapshot across every source and runs it through a staleness gate: if a feed is late, silent, or contradicts itself, Cintrel flags the gap rather than quietly reasoning on bad data. Synthesis runs against a memory of past cycles, so trends are visible, not re-derived.

Scheduled snapshotStaleness gateMemory of prior cycles
3
Decide

A deterministic rules engine plus one language-model pass ranks what changed by leverage, not recency. The output is not a wall of charts. It is a short, ordered list, with the top item framed as a decision: what changed, why it matters, the recommended call, with confidence shown and evidence linked.

Rules engineOne pass per cycleRanked by leverageConfidence + evidence
4
Act

The decision arrives with the work mostly done: a drafted reply, a stakeholder brief, a flow scaffolded from past campaigns, a commission flagged to reconcile. A human approves before anything ships. Cintrel removes the blank page, not the operator.

Drafts + briefsHuman approvesNothing auto-sent
Honest about what is live

Today Cintrel runs on a deterministic engine of seven rules plus one language-model synthesis pass per cycle, anchored by a persistent memory layer. The eleven named agents are how that capability is organized in the v1 spec, not eleven autonomous bots. They all run through that one shared engine. The same engine carries vertical rule packs: DTC live, restaurants built as a demo, and medical and dental, trades and field services, and auto service groups in pack spec. The full breakdown lives on the status page.

03 The surface area

Fourteen modules, one screen.

Each module exists because a working operator needed it that week. Together they cover the surface a consumer brand actually runs on.

01Acquisition
CAC, blended ROAS, and channel mix across Meta, Google, and organic.
02Customer service
Ticket volume, response time, and the themes driving contacts this week.
03Inventory
Stock cover by SKU, reorder timing, and the variants about to sell out.
04Marketing intel
Campaign and flow performance read against revenue, not vanity opens.
05Compliance
The regulatory and policy checks a consumer brand cannot afford to miss.
06VIP briefings
A per-stakeholder brief: what changed, what it means, what to do next.
07Competitive
Pricing, assortment, and positioning moves from the brands you watch.
08Reputation
Reviews, ratings, and sentiment pulled into one trend line.
09SEO + AEO
Search and answer-engine visibility, and the pages losing ground.
10Affiliates
Payouts owed, partners gone quiet, and commissions left unreconciled.
11Deploy health
Hourly checks on every page, API, and connector; a flag when one goes silent.
12Email flows
Flow coverage, the gaps, and the drafts that never shipped.
13Social
Organic reach and the content that actually moved revenue.
14Memory layer
Every investigation, decision, and result, kept so nothing gets re-litigated.
04 Under the hood

One cross-stack read, organized into named jobs.

The cross-stack capability above is organized internally, in the v1 spec, as eleven named jobs grouped by what they do. These are not eleven autonomous bots acting on your business on their own. They are how the one daily read is structured, and the honest live-versus-planned status of each is public on the status page.

Watches

It watches.

  • The Service Mind Engine
  • The Tracker Engine
  • The Forecaster Live
  • The Watchman Live
Writes

It writes.

  • The Author Engine
  • The Studio Planned
  • The Concierge Engine
Warns

It warns.

  • The Counsel Engine
  • The Auditor Live
Learns

It learns.

  • The Memory Partial
  • The Critic Planned
05 See it honestly

See exactly what is live today.

No vapor. The status page lists every module and agent with an honest state: in production, running on the engine today, or planned for the rebuild.