AFFENTRA

The operating system for affiliate businesses.

One operator. An entire AI‑powered affiliate company.

You set the objectives, the capital, the strategy and the risk boundaries. Affentra coordinates the research, structure, content, quality control and deployment around that direction — and brings you the small number of decisions where your judgment changes the outcome.

The Affentra home screen: the decision waiting for the operator above every figure, then counts for decisions waiting, runs in progress, runs needing attention and runs failed, then projects and sites, then the most recent run and what it is waiting on.
One operator's view of everything operating, in a demo workspace.Scroll to inspect the interface.

The control room

You direct the company. Affentra runs it.

An affiliate business is not one job. It is research, market selection, architecture, editorial, quality control, engineering, deployment and operations — and the coordination between them is what actually costs.

Affentra is the layer that holds those functions together. You state what you want and what you are willing to spend and risk; it plans the work, delegates it, executes it under policy, and records everything it did. You are the operator, not the person writing prompts.

  • A chatbot
  • An AI writing tool
  • A WordPress plugin
  • A collection of disconnected agents
  • Another SaaS dashboard

Affentra is the operating system that coordinates the whole affiliate company.

Why this has to exist

Affiliate businesses do not fail to grow. They get expensive to run.

Every new market adds research, structure, content, review, deployment and reporting. Historically the only way to absorb that is more people, more tools and more handoffs between them.

The work is not the hard part. Coordinating it is. Most of an operator’s day ends up spent moving work between systems rather than deciding anything.

How it scales today

  • Keyword and market research tools
  • Content briefs and editorial workflow
  • Writers, editors and reviewers
  • A CMS per property
  • Developers for structure and templates
  • QA checklists and spreadsheets
  • Deployment and release process
  • Reporting assembled by hand

How it scales with Affentra

An objective, a budget and a risk boundary

Affentra coordinates the work

Execution is recorded and reviewable

Human judgment where it changes the outcome

Everything else keeps operating

The operating lifecycle

One objective becomes coordinated work.

Affentra is not a tool you operate. It is a system that operates, inside boundaries you set, and returns to you when a decision matters.

  1. Objective

    You state the outcome and the boundaries

  2. Plan

    The work is decided before it starts

  3. Approve

    You decide anything consequential

  4. Execute

    Specialists carry it out under policy

  5. Deploy

    Quality-gated, versioned, released

  6. MeasurePlanned

    What the property actually earns

  7. DecidePlanned

    Invest, improve, scale, hold or stop

Operating todayPlanned — designed into the same system, not built yet

Where Affentra is today. Objective through Deploy is running: the screenshots on this page are captures of it. Measure and Decide are designed into the same system and are not built yet, and every section covering them is marked Planned. Affentra is in active development and preparing for private beta.

What it does today

From a sentence to a live property.

One objective. A plan you approve before anything runs. Work delegated across specialist capabilities. A result you can audit line by line, and a site that is live at the end of it.

The four screens below are one run, in order, in a demo workspace. Nothing in them is a mockup.

The Affentra objective composer, filled in: the objective reads “Launch a Swedish affiliate site about dog insurance”, the budget is twenty-five euros described as a hard ceiling, the project is optional and left for the agent to decide, and the Start objective button states the ceiling back.
One. The operator states an outcome and a budget — the one thing a human originates.Scroll to inspect the interface.
An Affentra plan waiting for approval, showing the objective, what the agent proposes, what done looks like, the two actions it would take with their risk levels, and the estimated cost, above Approve plan and Reject plan.
Two. The plan is decided before anything executes. Approving it is what lets it start.Scroll to inspect the interface.
The Affentra execution tree: Core delegating to Scout, Architect, Writer, Editor and Deployer, each with its own state and duration, beside the selected run's task, state, duration, cost and step count.
Three. The work is delegated across specialists, each accountable for its own part.Scroll to inspect the interface.
An Affentra site screen for hundforsakringar.se, showing the live address, that version 1 is live, when it was deployed, the rollback position, and a quality report of sixteen checks with one warning and none blocking.
Four. A live, versioned, quality-gated property — not a draft.Scroll to inspect the interface.

The organisation

The functions of an affiliate company, inside one system.

Not a tool each department uses. The coordination layer those departments existed to provide.

  • Market research
  • SEO strategy
  • Editorial
  • Engineering
  • Quality control
  • Deployment
  • Operations
  • Portfolio management
AFFENTRA

One operator

Scout, Architect, Writer, Editor and Deployer are functions, not personalities.

Each is a specialist capability with its own permissions, its own budget and its own accountable record of what it did. Core plans the objective and delegates it; the specialists carry it out; the operator is not asked to supervise any of them.

Affentra is not organised around chatting with individual agents. A command — issued in the product today, or through a planned channel such as WhatsApp — enters the same governed operating system, is checked the same way and is recorded the same way. The intelligence is visible in the work getting done.

An operator states an objectiveWith a budget and a risk boundary
CorePlans the work and delegates it
  • ScoutResearch
  • ArchitectStructure
  • WriterContent
  • EditorReview
  • DeployerRelease
Executive InboxOnly what crosses a decision boundary
The operator decidesApprove, adjust or decline
Affentra keeps operatingAnd records what it did

Infrastructure

It does not hand you content. It runs the property.

No CMS to administer. No plugin stack to keep compatible. No separate publishing pipeline to maintain alongside the thing that writes.

A site in Affentra is a structured model, not a folder of pages: routes, content blocks, offers and metadata the system can reason about. It compiles that into an immutable, content-addressed version, gates it behind quality checks, promotes a version to live, and can return to any previous one without rebuilding.

Structured, not authored into a theme

Pages are composed from a component model with required and allowed elements per page type, so a regulated comparison page cannot ship without what it must carry.

Versioned, not overwritten

Every build is an immutable bundle. Going live is promoting a version; going back is promoting an earlier one. Neither rebuilds anything.

Gated, not hoped for

Quality checks run against the built bundle — links, disclosures, canonicals, structure — and a blocking failure stops the promotion rather than reporting it afterwards.

Served by Affentra

The runtime serves the deployed version directly. Managed production hosting is part of the operating model and is not generally available yet.

The Affentra site delivery detail: the live version and its deployment, above the full quality report listing each check, its severity and its result.
The quality report the promotion had to pass.Scroll to inspect the interface.

Management by exception

Affentra works. You decide when judgment matters.

Executive Inbox is not a task list. It is the small set of decisions where a person’s attention has economic or operational value — and everything below that threshold simply keeps running.

An item is a brief, not a notification. It says what Affentra wants to do, which business it affects, why a person is being asked, what happens if it is wrong, whether it can be undone, and exactly what will run.

Approving is not instructing. What you reviewed is what runs: a decision cannot be re-planned into something else between the moment you grant it and the moment it executes — and the moment it is settled, the work carries on without you.

What Affentra wants to do
Deploy to production
Which business it affects
Nordic Pet · hundforsakringar.se
Why a person is being asked
HIGH action under balanced autonomy
What the consequence is
HIGH
Whether it can be reversed
Yes, with a recorded rollback plan
What actually runs on approval
Exactly what you reviewed
An Affentra decision brief for deploying a site to production, showing why a person is being asked, that nobody has estimated what the decision is worth, the project and site it affects, a HIGH consequence, that it is reversible with a recorded rollback plan, and the exact action that will run.
An Affentra decision brief for deploying a site to production, showing why a person is being asked, that nobody has estimated what the decision is worth, the project and site it affects, a HIGH consequence, that it is reversible with a recorded rollback plan, and the exact action that will run.
The decision quoted above. Approve, adjust the input, or decline with a recorded reason.Scroll to inspect the interface.

The foundation

Governance is not the product. It is what makes the product usable.

Nobody wants to buy governance. What they want is a system they can leave running — and that is only possible if every action passes the same checks before it runs, and none of them are advisory.

An agent that might spend money it was not given, or deploy something nobody reviewed, is a demonstration. Affentra was built governance-first so that the autonomy above it is worth having.

Scoped authority
Every capability is granted deliberately. Nothing can act outside what it was given.
Budgets held before work starts
Spend is reserved up front, so a limit can refuse an action rather than report it afterwards.
Risk boundaries
You set what runs unattended and what has to come to you. Consequence decides, not confidence.
Reviewed actions are frozen
What you approve is exactly what runs. An approval cannot be re-planned into something else.
Reversible by default
A high-risk action cannot be requested without a way to undo it.
A full audit trail
Who asked, what ran, what it cost, what it changed, and who decided.
Isolated by tenant
Every read and write is bound to one workspace, enforced in the database itself.
Safe to retry
A repeated instruction settles once. Nothing is executed twice because a connection dropped.

The record decides what happened.

The outcome of a run is derived from the platform’s own record of execution — which steps were planned, which completed, what was approved, what was spent — and never from the model’s account of itself. A model may write the closing report; it does not get to decide whether the work succeeded.

It is the difference between a log and a record, and it is the reason a run history is worth reading a month later.

An Affentra run outcome reporting Completed with gaps, one of two planned steps completed, and one gap explaining that a planned step was not completed because approval was declined — above a final report in which the model states the run completed perfectly with no gaps.
A real run. The platform reports the gap; the narrative underneath it claims there was none.Scroll to inspect the interface.

Command and controlPlanned

Your company, in your pocket.

The executive command line, in natural language. WhatsApp is planned as a command channel into the same Affentra system — not a second product, and not a chatbot bolted to the side of one.

Ask what is happening. Get a short operational answer. Receive the alert when something needs you, and decide it there. The permissions, the policy engine and the audit trail are the ones already described on this page: a decision taken from a phone is the same decision, recorded the same way.

You

How are the German projects doing?

Affentra

Fourteen properties are healthy. Two need attention. One is waiting on your approval for a content rebuild.

You

Show me the highest expected-return opportunity.

Affentra

Nordic EV charging. The plan is ready for review.

You

Approve it.

Affentra

Approved. What you reviewed is what will run. I will come back when it is live, or if anything blocks.

An illustration of intended behaviour, set in type rather than shown as a screenshot — because it is not built. Affentra has no conversational interface today; decisions are made in the Executive Inbox.

Commercial intelligencePlanned

The half that closes the loop.

Everything above produces properties. What turns a set of properties into a portfolio is knowing what each one earns, and acting on it.

The intent is to connect traffic, rankings, conversions, revenue, cost and experiments to the operating history Affentra already keeps — so the system can say which market deserves more capital, which property needs work, and which one should stop. That is the Portfolio Brain, and it is the part we have not built.

  1. ObservePlanned

    Traffic, rankings, conversions, revenue

  2. EvaluatePlanned

    What a property earns against what it cost

  3. LearnPlanned

    Carry the answer into the next market

  4. ImprovePlanned

    Change the property on evidence

  5. DecidePlanned

    Scale it, hold it, or stop it

One operator. An entire AI‑powered affiliate company.

Affentra is preparing for private beta. If you operate affiliate properties and want to run them this way, ask for access.

Already have access? Sign in