Cloud Browser Examples: 8 Business Workflows for AI Agents

At TinyFish, we help businesses run AI agents and cloud browsers in production, reliably and at scale. For example, DoorDash runs more than 1M web workflows a quarter with our cloud browsers, and Google Hotels uses TinyFish to reach more than 40,000 properties in Japan.

A question we hear often is: “What business workflows can an AI agent handle through a browser?”

So we looked at our customers' workflows to identify real-world examples.

These 8 cloud browser examples show how agents collect quotes, download invoices, check availability and complete other everyday business tasks. Each example explains what the agent does, what it needs from the browser and how teams run the workflow in production.

For a feature-by-feature comparison of providers, see our cloud browser comparison.

How cloud browsers and AI agents work together

Before we dive in, here is a quick reminder of how cloud browsers work.

A cloud browser is a real browser running on a remote server. In an AI agent browser workflow, the agent reads the page and decides what to do next, while the cloud browser executes its instructions. Scripts can also control the browser when the steps are predefined.

The workflow has three parts:

1. Your script or AI agent sends instructions, such as opening a page or filling in a form.
2. The cloud browser executes those instructions: it loads pages, runs JavaScript and interacts with the website.
3. The website responds, and your script or agent reads the result to work out the next step.

The provider keeps the browsers running so you can run several jobs at once. With TinyFish, TinyBrowser is the cloud browser and TinyAgent is the AI web agent built on top of it.

What AI agents need from a cloud browser

To complete business workflows, AI agents need reliable access to websites, the ability to repeat tasks at scale and results teams can check.

Look for these five capabilities in your cloud browser and agent setup:

• Logins. The most useful data sits behind an account. TinyAgent signs in with TinyVault, which fills in credentials from 1Password or Bitwarden inside the browser, and TinyProfiles preserves the session for later runs.
• Bot protection. TinyBrowser handles anti-bot checks and routes traffic through residential proxies at the infrastructure layer.
• Many different sites. One job often spans dozens of sites with different layouts, so sessions run in parallel.
• Pages that change. A script breaks when a button moves. TinyAgent works from a goal instead of fixed selectors.
• Results. TinyAgent returns structured results, with logs, screenshots and source URLs available for reviewing execution.

Each use case below needs a different mix of these.

1. Pull insurance quotes from multiple carrier sites

The first example on our list is collecting insurance quotes. An agent uses cloud browsers to enter the same customer details on several carrier sites and bring the quotes back for comparison.

TinyAgent takes the customer profile and a list of carrier sites, fills each form in its own TinyBrowser session and returns the premium, coverage limits, deductible and quote reference as structured JSON. Sessions run in parallel, so the wait depends on the slowest carrier. Ask for declines in the output too, so a missing quote doesn't look like a skipped carrier.

The Zebra runs the web steps of its insurance comparison workflow on TinyFish, with quote turnaround under two minutes in an always-on production deployment. "With TinyFish, workflows that once took minutes now happen in seconds. That speed compounds at scale," says Daniel Herrington, its Chief AI Officer.

QuoteSweep's CEO, Ankur Shrestha, describes the same job from the broker side: "The portal is the product. A model built to operate the live web is the difference between automation that demos well and automation that binds policies."

2. Download invoices and documents from supplier portals behind logins

Our next example is downloading invoices from supplier portals. An agent signs in through a cloud browser, finds new documents and downloads them for the finance team.

TinyAgent handles the sign-in through TinyVault. You connect 1Password or Bitwarden, and TinyVault fills in the credentials inside the TinyBrowser session, including the TOTP code when the vault item has one, so passwords never reach the model or your code. TinyProfiles preserves the signed-in session, so later runs restore it instead of logging in again, and several runs can work in the same account at once. When the session expires, TinyVault signs in again.

You ask for the fields your finance system needs, such as invoice number, date, amount and source URL, and each run returns them with the downloaded files and screenshots. Your system can then match the documents against purchase orders.

Teams already automate invoice downloads and document workflows on TinyFish across portals that require sign-in. Start with one portal you check every week, and add more once the output matches what you would have filed by hand.

3. Track prices and availability across competitor and long-tail sites

Price and availability tracking is another cloud browser workflow. Agents check competitor and supplier sites throughout the day, returning current prices and stock or booking availability.

Each check is a TinyAgent run on a TinyBrowser session. The agent opens the site, enters the dates or location, reads the price and availability and returns structured results. Runs go out in parallel and repeat as often as the market moves.

GetGo, a car-sharing platform in Southeast Asia, checks competitor availability and pricing several times a day with about 100% competitor coverage. Google Hotels uses TinyFish to reach more than 40,000 small hotels and guesthouses in Japan that standard booking integrations miss. That supply brings 3M+ annual impressions and a 20–30% increase in search visibility. Across our hotel work, we run hundreds of thousands of price and inventory checks a day across tens of thousands of properties.

At scale, DoorDash runs more than 1M web workflows a quarter on TinyFish across global markets, on authenticated systems and dynamic pages with repeated multi-step tasks.

4. Keep schedules and listings accurate across thousands of venues

Keeping venue listings accurate is another cloud browser workflow. Agents check live booking pages against published schedules so teams can catch changes before customers book.

ClassPass lists classes from more than 32,000 studios and gyms. Its staff used to sign in to studio systems, open booking interfaces, check class times and catch mismatches before members booked a class that had moved.

TinyFish agents now sign in to those studio systems, check live schedules, verify availability and flag discrepancies. ClassPass automated checks for 835 venues this way, cutting manual checking work by 98.6% and saving $43K a year.

Verification jobs like this need logins that work across hundreds of different systems, which TinyAgent handles with TinyVault and TinyProfiles. They also need results you can compare from one run to the next. Ask for stable identifiers, such as the class ID, date, time zone and start time, instead of a summary of the page. With stable IDs, a moved class shows up as a changed field, and a renamed class doesn't look like a new one.

The same setup fits store hours, menus, event listings and any directory your customers rely on. Start with the venues where mistakes reach customers most often.

5. Verify healthcare provider data and track policy changes

Our fifth example is verifying healthcare provider information and tracking policy updates. An agent uses a cloud browser when the source needs JavaScript, a login or several steps to reach.

Bonfire Analytics, a healthcare intelligence platform, matches messy CRM records to the right organization by finding each one's National Provider Identifier (NPI) on the web. CRM entries arrive with abbreviations, satellite clinic addresses and typos, and a wrong match attaches the wrong organization's data to an account. Bonfire runs millions of lookups a year through TinyFish and performs this matching during onboarding for every new customer. "Our healthcare provider intelligence rests on working the live web reliably and affordably," says Jaya Pokuri, Bonfire's CTO.

Most of this work starts with TinySearch and TinyFetch, which find and read public sources for free within published limits. TinyBrowser and TinyAgent take over when a source needs JavaScript, a login or several steps to reach. Monitoring CMS and Medicaid sources on a recurring schedule is one of the jobs teams already run on TinyFish.

6. Monitor competitor and regulatory changes

Monitoring website changes is another cloud browser workflow. An agent revisits competitor or regulatory pages, reads the relevant details and returns them for comparison with the last check.

For public pages, TinyFetch returns the page as clean content your system can compare with the last snapshot. When a page renders in JavaScript, sits behind a sign-in or takes a few clicks to reach, the same job moves to TinyBrowser and TinyAgent. Compare runs on stable fields, such as a plan name and its price, so a redesign doesn't read as a price change. Our web signal monitoring works this way.

Digital Garage, a Japanese company serving finance, retail and healthcare clients, monitors product launches, pricing, regulatory updates and market shifts across Japanese e-commerce sites and regulatory portals. Traditional tools couldn't navigate those sites, and manual checks couldn't keep up. "TinyFish gave us the ability to run web workflows we couldn't automate any other way in Japan," says Yuto Saeki, its Head of AI Strategy.

Also in Japan, we're working with NEC Networks & System Integration (NESIC) on an application that connects agents to more than 70 live sources, from earthquake sensors and weather feeds to supplier networks, for disaster preparedness and supply-chain monitoring.

7. Run browser tests at scale

Our seventh example is running browser tests. Test scripts or agents use cloud browsers to click through a site, fill in forms and check that each flow works.

TestSprite is an autonomous testing platform used by around 100K people. Its agents analyze codebases, generate tests, run them and report the results. Browser interaction is one of the execution layers those agents use, and TestSprite chose TinyFish for its reliability and precision at scale. "Testing at the speed of AI-generated code means driving real browsers reliably, thousands of times a day," says Yunhao Jiao, its CEO.

For QA, every session has to start clean, because a cookie left over from the last run can make a broken test pass. Each TinyBrowser session runs in its own microVM and is destroyed when the session ends. Your existing Playwright or Puppeteer tests connect to it over CDP and run as they are. Parallel sessions let a full suite run at once instead of one test at a time.

8. Build product catalogs and supplier data from many sites

The final example on our list is building product catalogs and supplier records. Agents use cloud browsers to collect details from sites with filters, pagination or logins, ready for your system to organize.

Ansur deploys AI workers into manufacturing companies. When a manufacturer needs a supplier that isn't in its database, the worker searches the web and keeps the workflow going. Ansur uses TinySearch to find supplier candidates and TinyBrowser for sources that only show their content inside a running browser. Switching from its previous provider took an afternoon. "At the end of the day it's an API switch, so there wasn't much standing in the way of trying it," says Ansur's founder.

For catalog work, define the schema before the first run: product name, manufacturer part number, attributes, price, availability and source URL. TinyFetch returns structured content from simple pages, and TinyAgent handles sites where products sit behind filters, pagination or a sign-in. Clean, structured output is what makes classification practical across many long-tail sites.

Deploy AI agents on cloud browsers with TinyFish

Give your AI agents the browser access they need to collect quotes, download invoices, check availability and keep records current. TinyFish brings search, extraction, managed cloud browsers and agent execution together on one platform, with one API key. Connect your own scripts or agents to TinyBrowser, or give TinyAgent a goal and a URL to handle the website steps.

Start with the parts your workflow needs:

• Find and read the right sources. TinySearch finds relevant pages, and TinyFetch turns their content into clean Markdown, HTML or JSON. It handles dynamic and JavaScript-heavy pages, giving your application useful context without the navigation, ads and repeated page elements.
• Run your own browser code on managed infrastructure. TinyBrowser provides cloud browser sessions you can control with Playwright, Puppeteer or CDP. We handle the browser infrastructure, session scaling, residential proxy routing and anti-bot handling, so you can focus on the workflow.
• Give an agent the website work. With TinyAgent, you provide a goal and a URL. It navigates pages, fills forms, works through filters and pagination, and adjusts its next step as the site changes. Its purpose-built web execution model handles those steps while your application stays in charge of goals, orchestration and business decisions.
• Work inside the accounts your business relies on. For authorized TinyAgent workflows, TinyVault and TinyProfiles work together to supply credentials securely and preserve login state. That gives your agent a path through supplier portals, account dashboards and other sites where the useful work starts after sign-in.

As the workflow grows, you can run independent jobs in parallel across sites and accounts. Results come back with logs, screenshots and source URLs, so your team can inspect what happened. The aim is to keep the whole job fast and reliable while its results stay fresh and accurate. Clean page content also keeps irrelevant tokens out of your model's context.

Start with one task you already know well: checking a price, verifying a class time or downloading an invoice. Define the result you need, including its source URL, and compare the first runs with your manual process. Once the output is dependable, expand to more sites and schedule recurring runs through your application.

Create your TinyFish account and start with TinySearch and TinyFetch, free within their limits. When your workflow needs TinyBrowser or TinyAgent, add funds to your prepaid Wallet. See our pricing page for current rates and limits.