Short answer: This WebRun review covers a browser agent that does your recurring web chores: it clicks and types inside Gmail, HubSpot, QuickBooks and Xero, at $20 to $200 a month and about 1 cent per completed task. For a solo owner, the $20 Starter plan is the sensible test bed, with a 7-day free trial. For anyone who needs an audit log, only the $200 Business plan includes one, and that changes the buying decision.
WebRun launched on 20 October 2026, so this WebRun review is written six days into the product’s public life. There are no independent reviews yet, no long-term reliability data, and no hands-on testing behind this article. What exists is the vendor’s own documentation, its launch announcement, and one independent newsletter that checked the same prices. Anything the vendor has not published is flagged as unconfirmed, not guessed.
What WebRun actually is
WebRun is an AI business automation platform whose agents automate the repetitive work a small team does every week, from answering customer emails to chasing overdue invoices. The agents work inside the tools the business already uses, on a schedule it sets, according to the company’s launch announcement of 20 October 2026. There are three ways to set an agent up: describe the job in plain language, record your screen while you do it once, or start from one of more than 14,000 ready-made workflow templates. The vendor also publishes guides for more than 500 apps.
The key distinction is how the agent does the work. It drives a real browser, clicking and typing the way a person does, so there are no APIs to connect and no code to write. Jobs run in the cloud on the schedule you set, so your own computer does not need to stay on. That is the whole pitch: tools that never got an API become automatable anyway. It also explains the product’s biggest weakness: a browser agent inherits every fragility of the websites it drives. Longer jobs get a worklist: give an agent 40 leads to follow up with, and it works through the list one item at a time, tracking what is complete, what comes next, and anything it skipped, with the reason.
The one number that decides it: the 1-cent task
WebRun’s pricing page is unusually honest about its billing unit, and that honesty is the best thing about the product. Every plan includes the full platform: all models, full speed, every channel (Telegram, Slack, WhatsApp), API access, webhooks, and the full integration library. Plans differ on allowance and scale, not on core capability. There is no per-seat pricing at all, which is a welcome contrast with the seat-stacking bills covered in our look at what AI tools really cost small businesses.
| Plan | Monthly price | Tasks included | Agents at once | Active workflows |
|---|---|---|---|---|
| Starter | $20 | About 2,000 | 3 | 50 |
| Pro | $100 | About 10,000 | 10 | 100 |
| Business | $200 | About 20,000 | 50 | Unlimited |
One task is one job the agent completes end to end: pulling a report, updating a record, sending a follow-up. Each task runs in an agent session that costs about 1 cent, and a job that does not finish in one session continues in the next, which costs another cent. The plan price is what you pay; the credit is what you get to spend. Starter is $20 a month and includes $20 of usage, roughly 2,000 tasks at about 1 cent each. Unused credit is not charged again, and extra usage is billed from pay-as-you-go credits, so nothing stops when you pass your allowance.
Here is the arithmetic the pricing page does not spell out, and it is our calculation: the tiers are not a volume discount. At 10,000 tasks, Starter costs $20 plus 8,000 overage tasks at 1 cent, which is $100, exactly the Pro price. At 20,000 tasks, Starter plus overage is $200, exactly the Business price. So you never pay a higher per-task rate by staying on Starter; you pay more only for the scale features that come with the higher tiers. Those features are real: maximum run length rises from 30 minutes on Starter to 2 hours on Pro and unlimited on Business; session history goes from 7 days to 30 days to 1 year; connected accounts per environment rise from 5 to 25 to unlimited; and storage per environment rises from 5 GB to 20 GB to 100 GB. Buy the tier for the feature you need, not for a cheaper task, because there is no cheaper task.
The approval system is the real product
Any tool that clicks around inside your business software lives or dies on its guardrails, and WebRun’s are the strongest part of the launch. You decide which websites an agent can visit and which actions it may take. Automation policies check each proposed click, keystroke and page visit before it happens, then allow it, block it, or pause for your approval. An allowlist can keep an agent to the domains you name, and the Strict Lockdown preset also blocks shell commands and file uploads. When an action needs approval, the run pauses and asks you in Telegram or WhatsApp, and every run is recorded. The check runs on the action itself, so a blocked action stays blocked even if text in an email or on a web page tries to talk the agent into it: a useful defence against prompt injection through a customer’s inbox. That approval-first design is exactly what a written AI usage policy should require of any agent touching company systems, and it is why WebRun is easier to govern than the shadow AI tools employees adopt without telling anyone.
But the guardrails are rationed by tier, and that rationing bites. Starter includes exactly 1 policy, Pro 10, and Business unlimited policies. A single policy has to cover every agent and every job on Starter: workable for one person with three similar jobs, thin for a team with different risk levels across sales, finance and support. Worse, the audit log exists only on Business. A team that needs to prove who approved what, months later, cannot get that from Starter or Pro at any price short of the $200 tier.
One question the vendor’s pages do not answer: what happens when a target site throws up a CAPTCHA or a bot challenge. The launch material says agents use saved browser logins in persistent cloud environments, which covers ordinary logins, but it says nothing about CAPTCHAs. Do not assume the agent sails through them. Put your first workflow on a site you control and watch the first ten runs before handing it a site that fights back.
Try the numbers on your own workload
The formula is our calculation from the vendor’s published rates: monthly bill equals the plan price plus any tasks above the allowance multiplied by $0.01; hours saved equals tasks multiplied by minutes per task divided by 60; the value of that time equals hours saved multiplied by your hourly rate; net benefit equals time value minus the bill. The plan prices and allowances are the vendor’s published figures; the minutes per task and the hourly rate are your own numbers, and the result is only as honest as they are.
Where WebRun beats the connector tools, and where it does not
Connector tools such as Zapier, Make and n8n move data through official APIs. They are fast, deterministic, and they break loudly when an API changes. WebRun drives the browser instead, so it can automate the long tail of tools that never built an API, or steps inside a tool that its API does not expose. A weekly job that lives half in a booking inbox and half in a scheduling page is exactly the kind of work connectors handle badly and a browser agent handles naturally. Our earlier comparison of the three connector platforms shows how their billing units differ, and WebRun’s 1-cent task is refreshingly simple next to Zapier’s task counting and Make’s credit arithmetic.
The trade runs the other way too. A browser agent is slower than an API call, costs a cent or more per job where an API call costs fractions of a cent, and breaks quietly when a website redesigns a page. A connector workflow fails with an error code; a browser agent can click the wrong button with confidence. That is why the first month with any browser agent should be a supervised trial, not a handover. If a job is high-volume, identical every time, and already covered by an official API, the connector tool is still the better buy. If the job is irregular, spread across tools, and currently done by hand in a browser, WebRun is the more honest fit. If you are still deciding what to automate first, our guide to automating work with AI tools is the place to start.
There is a related comparison worth making with screen-watching automation. IrisGo, reviewed here earlier, watches how you work on your own device; WebRun is the cloud cousin that runs on its own schedule. A tool that watches your screen sees everything you do; a cloud tool sees only the accounts you connect, and that boundary is a genuine advantage for a business owner wary of what an agent can observe.
Who should not buy WebRun
Do not buy WebRun for a one-off job. The product is built for recurring work on a schedule; a single migration, data cleanup or research project is cheaper done by hand or with a one-time tool. Do not buy Starter if you need an audit log, multiple risk policies, or runs longer than 30 minutes; those are Business and Pro features, and no amount of overage credit buys them. Do not buy it to automate a site whose terms forbid automation, or to scrape a competitor’s pricing behind a login; a browser agent does not make a forbidden action permitted, it only makes it faster. And do not hand it payments, refunds or payroll changes in its first month. Let it draft, let it pause for approval, and earn the right to act alone one workflow at a time. Small teams should also weigh the approval burden: every paused run waits in Telegram or WhatsApp until a human answers, so start with read-only and draft-only jobs, where approval is a quick glance.
What we cannot confirm yet
This is a six-day-old product, and the honest section of any WebRun review written now is the list of unknowns. No independent reviewer has published a hands-on test of WebRun. No reliability figures exist: no published success rate, no uptime record, no data on how often agents stall on changed pages. The vendor’s own worked example, a cleaning company completing 300 jobs a month, claims 43 hours of busywork replaced for about $8 of task credit; that is the vendor’s arithmetic at its own $30-an-hour assumption, not an independent measurement, and the $8 covers task credit only, not the $20 plan price. Read it as an illustration of the billing model, not as a promise of savings.
Two more unknowns deserve a line each. First, the pay-as-you-go credit price beyond the implied 1 cent per task is unpublished. Second, the vendor’s comparison with the enterprise browser agent Hark Handoff is the vendor’s own page, not an independent benchmark. The independent newsletter that confirmed the launch prices is The Early Finish’s 23 October roundup, which matches the vendor’s $20, $100 and $200 figures and the 7-day trial. None of this makes WebRun a bad product. It makes it an unproven one, and the correct response to an unproven product with a free trial is to run the trial on a real job and measure it.
The WebRun review verdict
WebRun is the clearest-priced browser agent a small business can buy right now: one billing unit, no per-seat charge, every feature on every plan, and a trial that costs nothing for seven days. Its approval system is thoughtfully built, and its pricing honesty is rare in this category. The case against it is youth, not design: no independent testing, no published reliability data, an audit log locked behind the $200 tier, and one policy on the $20 tier. Start on Starter, automate one draft-only job, measure the first month against the calculator above, and only then decide whether the agent has earned unsupervised work. A promising tool with a fair price and an unproven record, best bought as a trial, not a commitment.
Is WebRun free?
No. There is no free plan. Starter costs $20 a month after a 7-day free trial that needs no credit card. Pro and Business have no trial.
How much does WebRun cost per month?
Starter is $20 a month for about 2,000 tasks, Pro $100 for about 10,000, and Business $200 for about 20,000. Extra usage is billed from pay-as-you-go credits, with no per-seat charge on any plan.
What counts as one WebRun task?
One task is one job the agent completes end to end, such as pulling a report or sending a follow-up. Each task runs in an agent session costing about 1 cent, and a job needing a second session costs another cent.
Can WebRun send emails or pay bills without my approval?
Only if you allow it. A policy can block an action or pause the run so you can approve, reject or take over before anything happens, with the request arriving in Telegram or WhatsApp.
Is WebRun better than Zapier or Make?
It is different, not simply better. Zapier, Make and n8n move data through official APIs, which is faster and more reliable where an API exists. WebRun drives a real browser, so it can automate tools and steps that have no API, but it breaks when websites redesign their pages.
What happens if I exceed my monthly task allowance?
Nothing stops. Extra usage is billed from pay-as-you-go credits, so your agents keep running. The vendor does not publish a separate pay-as-you-go credit price beyond the implied rate of about 1 cent per task.
Shurah is the founder of AI Tools Daily, tracking pricing, licensing and policy changes across AI tools so readers can make decisions without wading through marketing claims themselves.
