Surfaces

A team runs on this, not one person who set it up.

Most tools in this category are a chat window, usable by whoever configured it and by nobody else, which is how a company ends up with a capability that leaves when one person does. This is a chat window, a management console and an API, and all three enforce the same per-person rules.

  • Same rules on all three
  • Nobody handles a credential
  • Revocable in one click

Who can reach what

Meta AdsGoogle AdsYour database
  • PriyaPaid media, all accountsAnalyst
  • SamBrand and organic onlyAnalyst
  • ContractorAccess ended FridayRevoked

Granted per person, per source, per destination. Taken away in one click.

The three

Each one exists because a different person needs it.

The AI client

Claude, ChatGPT, or any client that speaks the same protocol. Your team asks from the account they already use all day, and reaches only the sources they have been granted.

  • Nine ready-made playbooks appear as named commands
  • A new joiner runs the same audit the same way on their first morning
  • No new tool for anyone to learn, adopt or forget
You

/meta-audit last 30 days

Ranked by spend, with the waste named.

CampaignFlag
Prospecting, broadFatigue, week 4
Retargeting, 7dAudience overlap

A named command, so it runs the same way for everyone.

The management console

A web console on its own address, for whoever is accountable. Sign in by email link, because there is no password to store, lose or rotate.

  • Add a person, see who has access and when they were last active
  • Watch which sources are healthy and read the audit trail
  • Nobody handles a credential at any point, including the administrator

Sources

Meta AdsSearch Console
  • DanaLast active this morningAdmin
  • PriyaLast active 2 days agoAnalyst

Nobody handles a credential, including whoever is looking at this.

The API

The same access rules over plain HTTP, because not everything that needs your numbers is a person typing.

  • A spreadsheet refreshes itself
  • An automation triggers on a threshold
  • A notebook pulls the same governed data an analyst would get by asking

GET/v1/spend?window=last_30d

scopethe caller's own grant, nothing wider

  • sources3 of 11
  • may_sendfalse

Not a second way in. The same grant, answered over HTTP.

The access model

Two independent layers, and you own the first one.

This is the answer to the question that ends deals at this size, and it is unusually good: whether the connector exists in your workspace at all is your administrator's decision, not ours.

Layer one

Your AI platform

Whether the connector appears at all is decided in your own Claude, ChatGPT or Gemini administration console, by your own administrator, under your existing policy. Cartora is not part of that decision.

  • Your admin console
  • Your existing policy
  • We are not in it
Layer two

Cartora

Which sources a person reaches, which destinations they may send to, and whether they may send at all. Granted per person and revocable in under a minute.

  • Per person
  • Per source
  • Per destination
Together

Enforced separately

Someone who can see the connector but holds no grant gets nothing at all, rather than a partial view. Most vendors cannot say the customer controls the first layer outright. Cartora can.

  • No partial view
  • Five roles
  • Read-only roles cannot send

Day one

Nine playbooks ship with it, so nobody faces a blank prompt.

They appear inside the AI client as named commands. A new team member runs the same audit the same way as the person who has been here a year.

  • Meta audit: ranks campaigns by spend and flags waste, fatigue, fragmentation and concentration
  • Google waste check: zero-conversion campaigns, underpacing, low click-through, duplicates
  • Optimization review: pending platform recommendations and what changed recently, as one prioritised list
  • Cross-channel CAC: blends channel spend with your database's registrations. The reconciliation, as a command
See the other 5 of 9
  • Weekly brand report: the standing summary across channels and your own database
  • Daily ops check: yesterday against the day before, anomaly flags, and a short needs-attention list
  • Creative audit: ranks active creatives on creative signals and names the winning pattern
  • Creative teardown: one video examined scene by scene, with a timestamped transcript
  • Creative brief: winning patterns turned into ready-to-shoot briefs, each citing its metric

Start here

See the console before you decide anything.

The fastest way to judge whether this is a team tool or a toy is to look at the screen where access is granted and taken away. We will show you that first, before any of the analysis.

Usually 45 minutes  ·  APAC (GMT+6), working across time zones