A HARPA AI alternative you set up in plain words

Updated

Choose BrowseWiz if you'd rather describe a recurring job in plain words than build it as a command. It fills a Google Sheet, refreshes it on a schedule and alerts you when a page changes the way you described. Stay with HARPA for checks every 15 seconds, email or webhook alerts, a lifetime plan or a lower price.

Add to Chrome – it's free

Track three competitors' changelogs without building a command

In HARPA, a weekly job like this means scheduling an AI command, and its guide says scheduling "requires a basic understanding of Custom Commands & Parameters" Source 1. In BrowseWiz, you describe the job, approve the plan and say when to repeat it:

What you type

Open the changelogs at tallyhq.example/changelog, bramble.example/releases and quillstack.example/news. Put each entry from the past week in a new "Competitor releases" sheet: date, what shipped, competitor, type (feature, fix or pricing) and a link to the entry.

Approve the plan, then grant tab access and sign in to Google once when asked. The first run creates the sheet in your Google Drive while you watch. Then reply "Repeat this every Friday at 9:00 and add only entries that aren't in the sheet yet" and approve the new plan.

The sheet after two Friday runs

The first run, in chat on Friday, Sep 25, added three rows. The next Friday's scheduled run added only the two entries posted since, including Bramble's price rise. Each link opens the entry it came from.

Example: Competitor releases (fictional companies)
DateWhat shippedCompetitorTypeLink
Sep 22Shared dashboardsTallyFeaturetallyhq.example/changelog/118
Sep 23Duplicate Gmail contacts fixedBramble CRMFixbramble.example/releases/4.12
Sep 24Bulk export to CSVQuillstackFeaturequillstack.example/news/csv
Sep 29Starter plan from $15 to $19Bramble CRMPricingbramble.example/releases/4.13
Oct 1Slack alerts for scheduled reportsTallyFeaturetallyhq.example/changelog/121

BrowseWiz vs HARPA AI

BrowseWiz vs HARPA AI. As of . The numbers link to the sources.
FeatureBrowseWizHARPA AI
Setting up a recurring job Source 1 Source 2Describe it in chat, run it once, then approve a scheduleSchedule an AI command, such as one you build from steps like NAVIGATE, CLICK and EXTRACT
Scheduled runs Source 3Free: 1 task, up to once a day. Pro and Max: 10 tasks, up to hourlyOn the Pros and Life plans
Monitor condition Source 4A sentence you write; a model judges each changeA trigger you pick: text appears or disappears, a price drops, the page content updates and others
Fastest check Source 4About every 5 minutesEvery 15 seconds; every 3 hours by default
Alerts Source 4Side panel notice, page pill or system notificationNotification, email, sound, webhook or Make.com
Google Sheets Source 4 Source 5 Source 6Creates the sheet in your Google Drive and keeps writing to it; can't edit a sheet you already haveAI reads, edits, clears and inserts rows (version 14.0); monitors send data through Make.com
Outside automation Source 5 Source 7Calls your APIs, webhooks and MCP servers as tools; no public API to start a runGRID API starts commands from Zapier, Make.com, n8n or REST; MCP tools too
Models Source 3One built-in model, BrowseWiz Chat, which schedules and monitors use; any OpenAI-compatible model with your own key, in chatGPT-5.4, Sonnet 4.6, Gemini 3.1 Pro and more; your own API keys, OpenRouter, free web sessions
Where it runs Source 4 Source 8Your Chrome; schedules and monitors run only while it's openIn your browser: Chrome, Brave, Opera and other Chromium browsers
Free plan Source 3$0: 10M credits a month, 1 daily scheduled task, 1 monitor$0: 10 web-session messages a day, up to 100 AI command runs in total, commands and web monitors
Paid plans, billed monthly Source 3Pro: $29 for 250M credits a month. Max: $79 for 1B credits a monthEveryone: $12 for 5 Megatokens a month. Pros: $19 for 15 Megatokens a month
Billed annually or once Source 3No annual or lifetime planEveryone: $144 a year for 75 Megatokens. Pros: $228 a year for 225. Life: $240 once for 30 Megatokens that don't renew

HARPA runs the steps you build. BrowseWiz works out the steps.

A custom HARPA command is a chain of steps, such as NAVIGATE to a URL, CLICK an element and EXTRACT its content Source 2. Steps that click or extract find their element with a selector you grab from the page or write in YAML Source 9. A HARPA monitor is configured too: you set the page, a trigger and the actions it fires Source 4.

In BrowseWiz, the description is the setup. The agent opens each page, works out where the data is and writes the rows. You watch the first run in chat, then approve a schedule. After that, scheduled runs approve their own clicks, and a safety check refuses any step that would send, pay, delete or submit something you didn't ask for. A monitor needs only a sentence, and a model decides whether a change matches it.

The trade-off is exactness. A HARPA command repeats the same steps on every run, and its trigger fires on a precise rule, as often as every 15 seconds. The BrowseWiz agent may take a different path from one run to the next, and a model can misjudge a change. If the job must repeat identical steps, build it in HARPA.

How it works

  1. Name the pages and the columns

    List the changelogs, release notes or pricing pages, and say what each row should hold. There are no steps, selectors or parameters to set.

  2. Approve the plan, then the schedule

    BrowseWiz shows its plan before it browses. After you grant tab access and sign in to Google once, the first run creates the sheet while you watch. Ask it to repeat every Friday at 9:00 and approve the new plan. That saves a scheduled task and runs nothing yet. Free allows up to once a day; Pro and Max, up to hourly.

  3. Let each run add the new rows

    Each scheduled run opens the pages in the BrowseWiz Tabs group in your signed-in Chrome, adds only the new entries to the sheet it created and alerts you when it finishes. See how the browser agent works.

Which one fits

Choose HARPA AI if…

  • You need checks faster than every 5 minutes, down to 15 seconds.
  • You want email or webhook alerts, or runs started from Zapier, Make.com, n8n or an API.
  • You want exact steps that repeat on every run.
  • You use it mostly as a sidebar to chat with pages, summarize videos and write, with 100+ ready-made commands Source 10.
  • Price matters most: $12 or $19 a month, or $240 once.
  • You want a choice of models, your own API keys or free web sessions, or Brave or Opera.

Choose BrowseWiz if…

  • You'd rather describe a recurring job than build a command and pick elements.
  • You want an alert when a change matches what you wrote, not a fixed trigger.
  • You want the combination: plain-English requests in your signed-in Chrome, rows in Google Sheets, scheduled runs and monitors while Chrome is open, and a plan you approve before it acts.
  • You want a free plan that includes a scheduled task and a monitor.

Where BrowseWiz falls short

  • Slower checks. Monitors check about every 5 minutes, with no faster setting.
  • Chrome must be open. Scheduled runs and monitors stop while Chrome is closed or your computer is off. For checks around the clock, use a cloud monitor.
  • Signed out means no data. If a site signs you out, scheduled runs land on its sign-in page until you sign in again in Chrome.
  • No email or webhook alerts. Alerts arrive in the side panel, on the page or as a system notification.
  • Its own sheets only. BrowseWiz creates the sheet in your Google Drive and keeps writing to it, but it can't edit a spreadsheet you already have.
  • Not scripted. Runs can differ slightly. When the output must be identical every time, a step-by-step command is safer.
  • One built-in model for automation. Schedules and monitors always use BrowseWiz Chat. Your own key or an OpenAI-compatible model works in chat only.
  • Not for volume. Thousands of URLs, proxies and captchas need a dedicated scraper.

What it costs

HARPA's paid plans cost less: $12 a month for Everyone, $19 for Pros, or $240 once for Life Source 3. BrowseWiz Free includes a scheduled task, while HARPA's plan grid lists scheduling on Pros and Life only.

  • Free

    $0 / month

    • 1 scheduled task, up to once a day
    • 1 page monitor
    • 10M credits a month
  • Pro

    $29 / month

    • Up to 10 scheduled tasks, as often as hourly
    • Up to 30 page monitors
    • 250M credits a month
  • Max

    $79 / month

    • Up to 10 scheduled tasks, as often as hourly
    • Up to 30 page monitors
    • 1B credits a month

One run, in each tool's unit. HARPA's FAQ puts 1 Megatoken at roughly 100 command runs on its Cloud models; your own key and web sessions don't use them Source 3. So Pros' 15 Megatokens a month come to about 1,500 runs (15 × 100). In BrowseWiz, one reply that reads a 30,000-character page costs about 7,500 × 5 + 500 × 30 ≈ 52,500 credits, and an agent run costs more because it re-sends the chat at each step. A monitor check that finds no change makes no model call. Prices exclude VAT. See how credits are counted.

Questions

Try it on the pages you check every week

Add BrowseWiz to Chrome and describe the job in plain words. The free plan includes one daily scheduled task and one monitor.

Add to Chrome – it's free