iterationlayer.com

Task Machine is the product that replaced Iteration Layer. It turns recurring work — follow-ups, reports, and small jobs that repeat every week — into AI agent workflows you can schedule, approve, and audit. It is built by the same team, but it is a different product solving a different problem, not a rebrand of Iteration Layer.

What happened to Iteration Layer

Iteration Layer was European AI worker infrastructure. According to the site, the team wound it down as a standalone product because running two things at once meant doing neither well, and their attention had moved elsewhere. The practical consequence for existing users: Iteration Layer's APIs are no longer available. If you arrived looking for that product, there is nothing to sign up for — the worker infrastructure idea now continues inside Task Machine.

What Task Machine actually does

The core idea is narrow and specific: take work that never really ends and hand it to AI agents without leaving the loop yourself.

• Controlled workflows — a repeated job becomes steps with owners, schedules, approvals, checks, and retries, so it runs the same way each time.
• Inbox-first control — approvals, questions, blockers, and finished results collect in one place, so you don't have to watch every run.
• Bring your own model subscription — agents run on subscriptions you already pay for, such as ChatGPT or GitHub Copilot.
• Audit trail — every decision, check, retry, tool call, and cost stays attached to the run.
• Spending limits — the site explicitly frames this as preventing "runaway work."

What kind of work it's for

The site names the target directly: the follow-ups, the reports, and the small recurring jobs that quietly eat your week. The pattern to look for is work that is repetitive, scheduled, and checkable — something you'd otherwise do the same way on a recurring basis.

It is less obviously suited to one-off, judgment-heavy tasks where there's no repeatable shape to encode.

How it differs from a typical automation tool

The distinguishing claim is control rather than full autonomy. The stated point is "not to take you out of the loop" but to give time back "without giving up the wheel." That shows up in the approval steps, the inbox as the control surface, and the per-run history — you can verify what happened instead of trusting a black box.

Getting started

The site points to Task Machine with a 7-day trial. Pricing details live on the Task Machine pricing page rather than on this site, so check there for current terms.

Yes. According to the product site, Task Machine agents can run on model subscriptions you already pay for, such as ChatGPT or GitHub Copilot. That means you don't have to buy a separate model subscription just to use Task Machine. The exact list of supported subscriptions and the setup steps live on the Task Machine site, so check there before assuming your specific plan is covered.

What "running on your existing subscription" means

Task Machine is built for recurring work — follow-ups, reports, and small jobs that repeat. You turn those into agent workflows, and the agents execute the runs.

The relevant claim from the site is that you can "run agents on the model subscriptions you already pay for, like ChatGPT or GitHub Copilot." So the model access comes from a subscription you hold, rather than from a new model plan bundled into Task Machine.

What you still get per run

Even when the model access comes from your own subscription, Task Machine attaches the run itself to its own controls and record-keeping:

• Assigned tasks — each run takes the tasks assigned to it.
• One audit trail — every decision, check, retry, tool call, and cost stays attached to the run.
• Control features — shared tasks, scheduled runs, approvals that arrive in your inbox, and spending limits.

So the subscription question and the control question are separate: your subscription supplies model access, while Task Machine supplies the workflow, approvals, and history around it.

What this does not tell you

The site does not publish which specific subscription tiers or plans are supported, nor the setup steps for connecting them. Those details are on the Task Machine site. If your plan matters to you, verify it there rather than assuming any ChatGPT or GitHub Copilot plan will work.

Background: why this is a Task Machine question

Iteration Layer, the earlier European AI worker infrastructure product, has shut down as a standalone product — its APIs are no longer available. The same team now builds Task Machine, a different product aimed at recurring work. If you arrived looking for Iteration Layer, the subscription question above applies to Task Machine, not to the old product.

Practical next step

If your goal is to avoid paying for a second model subscription, the path is:

1. Identify which model subscription you already hold (for example, ChatGPT or GitHub Copilot).
2. Check the Task Machine site for whether that subscription is supported and how to connect it.
3. Start with a recurring job you keep putting off, and confirm the run shows up with its audit trail and any approvals routed to your inbox.

If your subscription isn't supported, or the setup isn't documented, that's the point to ask Task Machine directly rather than assuming a workaround.

Iteration Layer has shut down as a standalone product, and its APIs are no longer available. If you came for Iteration Layer, the replacement is Task Machine — a different product built by the same team. To start, go to the Task Machine site and begin a 7-day trial; pricing details live on the Task Machine pricing page. There is no migration path for Iteration Layer APIs, so any integration you built on them needs to be replaced rather than ported.

What actually changed

This is not a rebrand. The team states plainly that Task Machine is a different product solving a different problem:

The team's own framing: they wound Iteration Layer down because trying to do two things at once meant doing neither well, and the work they kept thinking about was Task Machine.

Who Task Machine is for

It targets work that never really ends — the follow-ups, reports, and small recurring jobs that quietly eat your week. You hand those to AI agents and stay in the loop rather than stepping out of it.

Concretely, the product describes:

• Controlled workflows — turn a repeated job into steps with owners, schedules, approvals, checks, and retries, so it runs the same way every time.
• Inbox-first control — approvals, questions, blockers, and finished results gather in one place, so you don't have to watch every run.
• Your existing subscriptions — run agents on model subscriptions you already pay for, such as ChatGPT or GitHub Copilot.
• Audit trail — every decision, check, retry, tool call, and cost stays attached to the run.
• Spending limits — described as a guard against runaway work.

How to get started

1. Go to the Task Machine site. This is the only entry point described; there is no Iteration Layer product left to log into.
2. Start the 7-day trial. The trial is offered directly from the site.
3. Check pricing before you commit. The pricing page is linked from the Task Machine site — read it there rather than assuming trial terms carry over.
4. Pick one recurring job to hand over first. The team's own suggested starting point is "the work you keep putting off" — a single recurring task is a better first test than a whole workflow library.
5. Set your approvals and spending limits up front. Both are part of the control model, so configure them before the first scheduled run rather than after.
6. Review the run history. Each run leaves an audit trail; use it to confirm the agent did what you expected before you widen scope.

Common sticking points

• Expecting an API. Iteration Layer's APIs are gone and Task Machine is not presented as an API replacement. If your use case was programmatic infrastructure, this is a different shape of product — evaluate it on its own terms.
• Assuming the trial removes the decision. The trial is 7 days; pricing is a separate page. Check it before you build dependencies on the trial.
• Turning on too much at once. The control features — approvals, limits, audit history — only help if you actually configure them. Start narrow.
• Treating it as a rename. The team explicitly says it isn't. Anything you knew about Iteration Layer's behavior doesn't transfer.

If you need the old Iteration Layer, it isn't coming back — the team's stated position is that they went all in on Task Machine instead. The practical next step is the trial, one recurring task, and a look at the pricing page before the 7 days are up.

Task Machine keeps you in control by turning recurring work into structured agent workflows with human checkpoints, rather than handing a task to an AI and hoping for the best. Approvals, questions, blockers, and finished results all collect in one inbox; spending limits, checks, and retries are built into the workflow itself; and every decision, tool call, retry, and cost stays attached to the run as an audit trail. This applies if you have recurring work — follow-ups, reports, small jobs that quietly eat your week — and want AI agents to run it without giving up the wheel.

The core idea: you stay in the loop by design

Task Machine is built by the same team behind Iteration Layer, a European AI worker infrastructure product that has been shut down as a standalone offering. Its APIs are no longer available. Task Machine is a different product solving a different problem, and its stated point is not to take you out of the loop but to give you your time back without giving up the wheel.

That translates into four concrete control mechanisms.

1. Controlled workflows instead of open-ended prompts

A repeated job becomes a series of steps with defined owners, schedules, approvals, checks, and retries. Because the structure is fixed, the job runs the same way every time rather than depending on how a single prompt happens to land.

2. Inbox-first control

Approvals, questions, blockers, and finished results gather in one place. You don't have to watch every run or hunt through logs to find what needs you — the things that require a human decision come to you.

3. Spending limits and no runaway work

Spending limits are part of the workflow setup, and the product explicitly frames this as preventing runaway work. You set the boundary before the agent runs, not after you notice the bill.

4. A full audit trail per run

Every decision, check, retry, tool call, and cost stays attached to the run. Instead of trusting a black box, you can see exactly what happened. Each run takes assigned tasks and leaves one audit trail.

What this looks like in practice

Using subscriptions you already pay for

Agents can run on model subscriptions you already pay for, such as ChatGPT or GitHub Copilot. This matters for control in a practical sense: you're not adding a separate model billing relationship to manage, and the run still produces a single audit trail regardless of which subscription powers it.

Getting started

The site points to Task Machine with a 7-day trial. If you came looking for Iteration Layer specifically, note that its APIs are no longer available — Task Machine is the successor effort from the same team, not a continuation of the same product.

A reasonable first task is the one you keep putting off: a recurring follow-up, a recurring report, or a small recurring job. Set it up as a workflow with an owner, a schedule, and an approval step, then let the inbox tell you when it needs you.

Where the limits are

The available material describes the control model but does not specify pricing beyond a link to a pricing page, nor does it detail which models or integrations are supported beyond the examples of ChatGPT and GitHub Copilot. If your decision depends on exact cost, seat limits, or a specific integration, check the pricing page and current product documentation before committing.

Iteration Layer no longer exists as a standalone product. It was a European AI worker infrastructure project, and the team behind it has shut it down and moved on to build a different product called Task Machine. If you arrived at iterationlayer.com looking for the original service, the key facts are: the APIs are no longer available, and the site now functions as a handover page pointing visitors to Task Machine.

What Iteration Layer was

Iteration Layer was European AI worker infrastructure — a product the team describes as something they cared a lot about and poured months into. The site's own account is that the deeper the team got into it, the clearer it became that their attention was elsewhere.

That "elsewhere" turned out to be Task Machine, and the team made what they call the hard call to wind Iteration Layer down rather than split their focus.

What happened to it

Iteration Layer has been shut down as a standalone product. The team is explicit that this was not a quiet rebrand or a fresh coat of paint:

• The APIs are no longer available. If you were integrating with Iteration Layer, that integration no longer works, and the team acknowledges the disruption this causes.
• Task Machine is a different product, built by the same team, solving a different problem.
• The reasoning given is that trying to do two things at once meant doing neither of them justice, and the work the team couldn't stop thinking about was Task Machine.

What Task Machine is instead

Task Machine targets recurring work — the follow-ups, reports, and small repeating jobs that eat into a week. The pitch is that you hand these to AI agents while staying in control. The site lists these concrete features:

The stated goal is not to remove you from the loop but to give you time back without giving up the wheel.

If you came here for Iteration Layer

There is no migration path described on the page — the APIs are simply gone, and the team offers an apology for the disruption rather than a replacement for the old infrastructure. What they offer instead is Task Machine, which they describe as the thing they believe in enough to bet everything on.

If you want to evaluate that replacement, Task Machine offers a 7-day trial, and pricing details live at taskmachine.io/pricing. Whether Task Machine fits your use case depends on whether your work looks like the recurring, approval-heavy jobs described above — it is not a drop-in substitute for the AI worker infrastructure Iteration Layer provided.