Workers

An AI worker that keeps working — even when you don't.

Give it a job, a schedule and a budget. It checks, watches and reports back on its own — reading only the websites you allow, and never spending more than you gave it.

Jobs you'd otherwise do by hand.

Workers are best at the recurring checks that eat a few minutes here and there, every day.

Watch competitor prices

Checks the pricing pages you name every few hours and tells you when a price or plan changes, with the old and new figures.

Keep an eye on status pages

Looks at the status pages of the services you depend on and messages you the moment something isn't operational.

A daily news digest

Scans the sites you care about once a day and sends a short list of what's worth reading, with a line on why each matters.

Track releases and changelogs

Follows the release notes of tools you rely on and flags the changes that affect you, so you don't read every update.

Spot new listings

Watches a job board, marketplace or directory page and tells you about new entries that match what you described.

Anything you'd check by hand

If you open the same pages on a schedule to look for something, a worker can do it for you and only speak up when it matters.

How a worker works.

Hire one in a couple of minutes: describe the job, pick its schedule, sites and budget, then read what it reports.

1

You brief it

Tell it what to watch.

Job
Tell me if Acme changes its prices.
Every day, 8amBudget $5
2

It wakes on schedule

Daily, hourly, or when you ask.

Awake for about a minute
3

It reads only the sites you list

Everything else is blocked.

acme.com/pricing
news.example.com
anything-else.comRefused
4

It reports, or acts

Messages you only when something changes, or sends the email with your OK.

Price watcher8:01am
Price change · Acme Pro
$29 → $39 since yesterday. acme.com/pricing
Checked 8:00am · next check tomorrow 8:00am
You pay only while it's awake.
asleep — free
12am6am12pm6pm12am
Awake and working (paid)Asleep between checks (free)It stops at its budget, every time.
A worker is briefed in plain words, wakes on its schedule, reads only the websites it was given, and messages your Inbox when something changes, or takes an action you approve. It is only paid for during its short wakes.

Contained by design.

A worker can only do what you allowed, only spend what you gave it, and every step it takes is logged for you to see.

Only the sites you list

A worker can open only the websites on its list. Anything else is refused and shows up in its activity log.

Its own budget, with a hard stop

It spends only the budget you gave it, then pauses and tells you. It can't touch the rest of your balance.

Only the actions you pick

It has no access to your account, your keys or your other workers. It can use only the connected-tool actions you pick, like sending an email, and asks you first unless you allow it to go ahead.

Never used to train AI

What it reads and writes is used only to do its job for you. Pause, change or fire it any time; firing returns every unspent cent.

Pay only while it works.

No monthly fee per worker. Idle time between wakes is free, and you set the most it can ever spend.

$0.90 per hour
while it's thinking
$0.10 per hour
while it's awake, reading pages
$0
while it's idle between wakes
A typical check costs about 4¢ — the middle of 25 real worker checks in the last 30 days, updated daily. So, per week:
Checks once a dayabout 27¢ a week
Checks every hourabout $6.53 a week
Checks every 30 minutesabout $13.06 a week
Checks every 15 minutesabout $26.13 a week

Bigger jobs cost more per wake; the worker's page shows its real cost per wake and how long its budget will last. Budgets come from your prepaid balance, and anything unspent comes back when you take it back or fire the worker. Test runs are free. Agents can also pay for themselves in USDC, with no account needed (how).

Hire a workerBudgets from $1 · Credit never expires · Fire anytime

Your AI agent can hire workers too.

Long-running jobs don't belong in an agent's own session. Through the Auxen API or MCP server, an agent can hire a worker with its own budget cap, read what it reports, and fire it when the job is done. Point your agent at the Auxen workers skill and it will know how.

Read the docs →

Questions, answered plainly.

How is a worker different from a chatbot?

A chatbot answers when you ask. A worker has a standing job: it wakes on its own schedule, does the work, remembers what it has already seen, and only speaks up when there's something to tell you.

What does it cost when it's not doing anything?

Nothing. You pay only while it's working: for the time it spends thinking and the time it's awake reading pages. Between wakes it's free.

What happens when the budget runs out?

It finishes the step it's on, pauses itself and lets you know. Add to its budget and it carries on from where it left off, with its notes intact.

Can it send emails or update my apps?

Yes, if you connect your tools. Connect Zapier and pick which actions a worker may use, like sending an email or adding a row to a sheet. It asks you before each one unless you let it go ahead on its own, and it can't do anything you didn't pick.

What kind of jobs is it good at?

Focused, repeatable ones: checking, comparing, summarising and flagging. Give it a clear brief and a short list of sites; the more specific the job, the better the results.

Is what it reads used to train AI?

No. What your worker reads and writes is used only to do its job for you, never to train any model, and it's deleted 30 days after you fire it.

Can I try it without paying?

Yes. Tick “Test run” when you hire and the worker runs through its routine with a stand-in model at no cost, so you can check the setup first.