• Explainer
  • E-commerce
  • 5 min read

The Integrations Page, Explained: Write-Only Secrets, OAuth, and Adapters

One page, one health state per system, and a secret you can set but never read back.

Author
ftes.ai Editorial
ftes.ai Team

The short answer

The Integrations page in the commerce engine is where every connected system lives: the read-only commerce platform, the helpdesk, returns, warehouse, people counting, email marketing, behaviour analytics, search console, social and the AI providers. Secrets are write-only, connections use OAuth where available, and an adapter sits behind each capability.

Put a Digital FTE on thisEvery role proposes with evidence and reports to you.

Key takeaways

  • Secrets are write-only: they can be set and replaced, never read back out of the platform.
  • Each system shows its own health and its last successful run, linked to the run ledger.
  • Integrations are described by capability, and an adapter sits behind each one.
  • AI providers are cards too, so text, image and video vendors are configuration rather than code.
  • Adding a system is incremental: one at a time, each one earning its place.

What is on the Integrations page?

Every capability the engine can connect: the commerce platform, read-only and feeding everything; a helpdesk; a returns platform; warehouse and packing; people counting; email marketing; web behaviour analytics; search console; social platforms; outbound email for manager messages; and the AI model providers.

Each card shows whether it is connected, its health, its last successful run and what it contributes. A brand can see at a glance which parts of the picture are live and which are missing, which is also why the surfaces can say Not reported honestly.

  • The commerce platform: read-only, required, and the source everything else is joined to.
  • Helpdesk, returns, warehouse and packing, people counting: the operational systems.
  • Email marketing, behaviour analytics, search console and social: the growth systems.
  • Outbound email for manager and supervision messages.
  • The AI provider cards: one default each for text, images and video, swappable at any time.

Why secrets are write-only

An API key that can be read back out of a platform is a key that can leak from it. On this page a secret can be entered and replaced but never displayed, so the platform holds what it needs to work and cannot hand it to anyone, including a user who asks.

Where the system supports OAuth, that is used instead, so the brand grants a scoped connection it can revoke from its own side without touching the engine.

Rotating a secret is the same action as setting one: paste the new value and the old is replaced, without either ever being displayed. Where a provider supports scoped tokens, the page asks for the narrowest scope that supports the capability.

Capability, not vendor

The page is organised by what a system contributes rather than by brand name: tickets, return records, packing performance, footfall, reach, insights. That framing is deliberate, because it makes the architecture obvious: the engine needs a capability, and a specific vendor is one way to provide it.

Behind each capability sits an adapter. That is why another commerce platform is a build rather than a rewrite, and why a brand switching helpdesk does not lose its support history's shape.

It also keeps the language on the page honest with a brand that has not chosen a vendor yet: the question is whether the brand wants footfall joined to sales, not whether it wants a particular counter.

AI providers are configuration

Text, image and video model vendors are cards on the same page: one default for each, swappable. The platform is vendor-independent by design, and the AI performance surface shows what each choice costs in latency, reliability and tokens.

For a brand, that means a provider outage, a price change or a policy change is handled by changing a card rather than by waiting for the platform to be rewritten.

Why add one system at a time?

The commerce platform is the first and the only required one; everything else earns its place. A brand with heavy returns connects the returns platform next. A brand with shops connects people counting. A brand pushing win-back lists connects the email tool.

Each addition unlocks specific surfaces and specific joins, and the page says which. That makes the rollout a sequence of small decisions rather than a project.

What connecting does not do

It does not grant write access to the store: the commerce connection is read-only and cannot be configured otherwise. It does not share data between brands. And it does not turn on unattended sending anywhere; that remains a separate, explicit switch per class of action.

On ftes.ai, the same integrations carry the roles' own connections, and a role that loses a connection says so in its Setup tab rather than working from stale data.

In what order should a brand connect systems?

Commerce first, because everything is joined to it. Then the system that covers the brand's largest known problem: returns for a brand with heavy returns, people counting for a brand with shops, the helpdesk for a brand drowning in tickets.

Growth systems last, because their value depends on the commerce joins already being in place. A search console connection is far more useful once articles are being published against a Brand Brain than it is on day one.

Frequently asked questions

Can I read a secret back?

No. Secret fields are write-only: they can be set and replaced, never displayed.

Is the store connection ever writable?

No. The commerce platform connection is read-only by design.

Why organise by capability?

Because an adapter sits behind each capability, which is what makes another vendor or platform a build rather than a rewrite.

Are AI vendors configurable?

Yes. Text, image and video providers are swappable cards on the same page.

Do I need to connect everything?

No. The commerce platform is the only required one; the rest are added one at a time as they earn their place.

  • E-commerce
  • Commerce engine
  • Digital FTE
Further reading

Keep going.

Related guides on hiring and managing Digital FTEs.

Ready to hire your first Digital FTE?

Pick a role, connect the tools it needs, and review its first proposals this week.