Skip to content
Anslation

How AnsBot works: record, review, run, check.

You build a bot by recording the task with the AnsBot Recorder, describing it for a draft, or adding steps by hand. AnsBot then runs those steps in a browser on its servers, by hand, on a schedule, by webhook or from a batch, and keeps the evidence of each run.

From a task to a run. Six steps, in order.

  1. Record it

    Do the task once in your own browser with the AnsBot Recorder for Chrome. Each click, choice and field becomes a step; what you type becomes an {{input}}.

  2. Or describe it

    Give a start URL and a goal. AnsBot drafts the steps from the page's structure and saves a draft that never runs on its own.

  3. Review the steps

    Check what each step opens, fills, clicks, checks or reads, and how it finds its element.

  4. Add inputs, sign-ins and approvals

    Name the inputs, pick a vault entry or browser profile, and mark the steps that must wait for a person.

  5. Test, then publish a version

    Run it in test mode, send the version for review if you want one, then publish it.

  6. Run it and read the evidence

    Start runs by hand, on a schedule, by webhook or from a batch, then read each run's log, output and screenshots.

The twelve step types. Every bot is built from these.

AnsBot step types
StepWhat it does
Go to URLOpens a web address; {{inputs}} can fill parts of it.
ClickClicks an element.
FillTypes a value, usually an {{input}}, into a field.
SelectPicks an option in a dropdown.
CheckTicks a checkbox or radio button.
UncheckClears a checkbox.
WaitPauses for a set time.
Verify textWaits until a text appears, and fails the step if it does not.
Extract textReads an element's text into the run's output.
Extract listReads a list of items, with title, text and link, into the output.
Extract tableReads a table's rows into the output.
ScreenshotSaves a full-page screenshot to the run.

Every step has its own timeout, can be optional, and can wait for approval.

What a bot can do, and what it cannot.

It does

  • Finds elements by CSS selector, role, label, placeholder or visible text, with fallbacks it has learned.
  • Takes inputs from the console, a schedule, a webhook body, a batch row or a client portal.
  • Reuses a signed-in session from a browser profile.

It does not

  • Upload or download files.
  • Branch or loop: steps run in order, and an optional step is skipped if it fails.
  • Run on cron expressions or in a chosen time zone; schedules are hourly, daily or weekly.
  • Run inside your own Chrome; runs use a headless browser on Anslation's servers.
  • Reach private or intranet addresses.

What every run leaves behind. Evidence you can read afterwards.

  • Step logEach step's status, the element it used, a message and its output.
  • OutputThe values extract steps read, under the keys you chose, with secret values redacted.
  • ScreenshotsAt screenshot steps, when a run waits for approval, when an operator pauses it, and when a step fails.
  • Failure traceA private browser trace of a failed or waiting attempt, for roles allowed to see it, kept for a limited time.
  • CheckpointWhere a failed run stopped, so you can resume from that step.

How it works: common questions.

Still have a question? Ask us

Do I need to write code to build a bot?

No scripts. You record the task with the AnsBot Recorder, describe it for a draft, or pick steps in the editor; a step finds its element by CSS selector, role, label, placeholder or visible text.

What does the AnsBot Recorder record?

Clicks, choices and typing in the tab you choose, turned into steps for one bot. It does not send what you type: each field becomes an input such as {{email}}, and password, hidden, file, card and one-time-code fields are skipped. It needs Chrome 120 or later.

Where does a run happen?

In a headless browser on Anslation's servers, not in your own Chrome. Each run opens a fresh browser; a browser profile can carry a signed-in session from one run to the next.

Can a bot run on a schedule?

Yes: hourly, daily or weekly, at an interval you set, each schedule with its own inputs and its own test or production mode.

What happens when a step fails?

When a required step fails, the run stops there, takes a screenshot and saves a checkpoint, so you can resume from that step once it is fixed. If a fallback target worked instead, the run carries on and proposes that target as a repair for review.

Have a task in mind? Ask for access.

Tell us the web task and the site it runs on. We read every request and reply by email.

Office hours: Monday to Saturday, 10:00–18:00 IST+91 74158 85838