August 4, 20266 min readBy Conto Team

The Web Is Full of Useful Work APIs Can’t Reach

Some of the most useful work on the internet still lives inside websites: comparing products, checking filings, finding current listings, monitoring pages, and collecting information from workflows that were never designed as APIs.

Browser automation can do that work, but a blank prompt hides what matters: which site, what the browser will click, whether it can change anything, and what the run costs.

The Enchant Browser Library turns those open-ended tasks into reviewed starting points. Each task explains what it does, what it needs, what result to expect, and what it costs before anything paid begins.

Why browser work matters

APIs are the best path when information is structured and directly available. A browser is the better tool when a task requires navigating a real site, interacting with a page, collecting information from a workflow, or working with a site that doesn't expose the right API.

A browser can be the only practical interface to a live marketplace, a public filing system, a news site, a job board, or a workflow that changes as the site changes. But that flexibility creates ambiguity. A request such as “research this company” could mean a read-only search, a login-protected workflow, a form submission, or a long browser session. Each version needs different inputs and carries different risks.

A browser library isn't just more automations. It gives people a way to use the open web without designing, paying for, and supervising each browser workflow from scratch.

From a blank prompt to a bounded task

Without a starting point, browser work usually means figuring out the site, designing the workflow, gathering the required inputs, writing or adapting automation code, setting up another browser provider, and hoping the cost and behavior are what you expected.

The Browser Library changes the starting point:

Choose a reviewed task that matches the job.
See the site or source, how the task works, and the expected result.
Know whether the task is read-only or may write.
Provide the details the workflow actually needs.
Review what will happen and the price before anything paid runs.
Receive the result and a receipt in the same workspace.

That turns browser automation from an opaque capability into an inspectable workflow.

Start from a reviewed task

The library includes presets for jobs such as product search, current weather, video transcript extraction, SEC filing research, job search, repository metadata, paper search, financial news, and other structured browser work.

Each entry gives you useful context before you run it:

The site or source the task will use
How the task works
Whether the task is read-only or may write
The details the task needs from you
The result format you should expect

Choose a preset and Enchant opens a structured draft in chat. Add the missing details, adjust the instructions, and review the task before it starts.

Presets and templates

The library has two kinds of starting point:

Browser presets are Enchant's reviewed task guides. They describe a specific job, tell you what information is needed, and explain what result to expect.
Browserbase templates are reusable starting points for browser workflows. They're useful when you want to begin with a broader workflow and adapt it to your situation in chat.

A preset is a defined task; a template is a flexible starting point. In both cases, Enchant shows the safety boundaries, what you need to provide, and the price before paid work begins.

Try these tasks

Start by opening the matching preset or template in the Browser Library, then add the details that matter to you:

“Find wireless noise-cancelling headphones under $300 on Amazon. Return the product title, price, rating, and canonical link.”
“Get the current conditions, seven-day forecast, and active alerts for San Francisco, California.”
“Extract the full timestamped transcript from this YouTube video: [paste a video URL].”
“Search SEC EDGAR for Acme Corporation’s recent 10-K and 10-Q filings. Return the filing dates, accession numbers, and document links.”
“Find the cheapest round-trip flights from SFO to JFK for September 12–16, including airline, stops, duration, price, and booking link.”

They show the kind of result to ask for. The library supplies the right workflow and questions; you supply the query, constraints, and output you care about.

A product search example

Suppose you want to compare products on Amazon within a budget.

Instead of starting with a vague request and hoping the workflow understands the scope, open the Amazon Product Search preset. Add the query, storefront, filters, and maximum number of results. Enchant checks that the request is complete, shows what will happen, and gives you an estimated price before the browser session begins.

After approval, the browser collects the requested information and returns the result in the same workspace as the request. You don't need to write automation code or create a separate Browserbase account for a one-off task.

You know what will happen, what the task needs, whether it can change anything, and what it'll cost before you authorize it.

The same pattern works for a fixed research workflow or a custom browser job. Start from a reviewed template when one fits. Describe the task in your own words when it doesn't.

Why this is more than a catalog

A list of browser skills is useful. The boundary around each task is what makes the list usable.

Each task in the library shows the site it uses, whether it's read-only, the details it needs, the expected output, and the price. The result comes back with a receipt.

That boundary matters for people, teams, and agents. A person can run a one-off task without learning browser infrastructure. A team can begin from a shared workflow instead of rebuilding the same research process. An agent can discover a task, understand what it needs, and carry the price and approval rules through to the run.

The infrastructure stays out of the way

Browserbase provides the browser infrastructure behind these tasks. Enchant handles the part people see: choosing a task, collecting the details, showing the boundaries and price, and returning the result.

The infrastructure can be powerful without making each user learn how it works. Enchant makes the boundary around the task visible instead of leaving it inside an open-ended prompt.

Safety stays visible

The library labels tasks as read-only or potentially write-capable. Enchant checks required details before a run and pauses for CAPTCHAs, sign-in requests, or anything it can't verify.

Paid browser work follows the same controls as the rest of Enchant. The selected task and estimated price are shown before it runs. Spending limits and approval rules can stop a request before credits are used.

You still see what the browser does. You also see the boundary before it runs: the site, read-only or not, and the price.