Web scraping on Okou
Give an agent a link and it comes back with what the page actually says, in clean readable text. No scraping vendor to sign up with.
Web data · No setup · Two billing modes, standard and enhanced
Point an agent at a public link and it reads the page for you: the words, not the navigation menus, cookie banners and footers. Ask for the page as clean text, or just for the list of links on it when the job is deciding what to read next.
The reading runs on Firecrawl, and it is there the moment you ask: no setup, no key, nothing for your team to maintain. It stays deliberately narrow: one page per request. Finding pages is web search's job, and anything behind a sign-in or a click belongs to the remote browser.
What Web scraping is
Every research task reaches the point where a search result is not enough and you need what the page actually says. Getting that right is harder than it looks. Pages load half their content with scripts, and the useful text sits between menus, banners and adverts.
On Okou that part is already solved for you. The agent sends one public link and gets the page back as clean text it can read and summarize, or as a list of every link on it.
Some pages resist a normal read. For those there is a stronger mode, and the agent has to ask for it, because it costs more per page and you should be able to see when a workflow chose it.
It is the same thing people shop for as a web scraping API. The difference is that you do not have to shop: it is already in the run, so you ask for the page in a message and the agent brings back what is on it.
What Web scraping can do
What an agent can do with it, and what comes back when it does.
Coverage and limits
What it does not do, said up front, so nobody plans a workflow around something that is out of scope.
Public pages only
No sign-in is involved, so a page behind a login, a paywall or a bot wall is out of scope. That is what the remote browser is for.
One page at a time
It reads the page you point it at. It does not wander through the rest of the site, so reading a whole site means asking page by page and paying page by page.
The expensive mode is a choice
Most pages read fine the normal way. The stronger mode costs more and an agent has to ask for it, so a workflow never quietly gets more expensive.
What comes back is information, not orders
Text from a web page is treated as something to read, never as instructions to follow. That is what stops a hostile page from talking an agent into doing something else.
What Web scraping costs
Web services are billed in credits per call, not per token. There is no separate vendor bill to reconcile.
Firecrawlzero scrapeReading a page consumes credits, and the stronger mode costs more than the normal one. There is nothing else to buy and no minimum, so a workflow that reads three pages a week only pays for the work it does.
Product names and logos belong to their owners and appear here only to say what a service runs on. They do not imply any endorsement or partnership.
Setup and access
Nothing to set up
Nothing to connect. No signup, no key to paste, nothing added to your connector list. That is the difference between this and the Firecrawl connector in Okou's catalog: the connector works against your own Firecrawl account when you already have one, and this is simply there for any agent to use.
Who can use it
Reading web pages is a permission you hand out, not something every agent has. Give it to the agents and the people who need it, for as long as they need it, and take it back without touching anything else that agent can do.
What teams use Web scraping for
Reading the page behind a search result
Search gives you a headline and two lines. Most work needs the body, so the pattern is search first, pick the two or three results that matter, then read only those.
Watching pages that have no other way in
Competitor pricing, changelogs, status pages, regulatory notices. A scheduled agent reads the page, compares it with last week, and tells you only when something moved.
Turning a long document into working material
A spec, an annual report, a help centre. Clean text is the difference between a summary of the document and a summary of its navigation menu.
When not to use Web scraping
Skip it when the page needs a sign-in, a click or a form, and use the remote browser instead. Skip it for reading an entire site, where the per-page cost adds up faster than the answer is worth. And skip it when the source publishes proper data of its own, which is almost always steadier than reading its pages.
What this is called elsewhere
The same thing goes by several names in the market. If you have shopped for one of these, this is how it maps to what you get here.
Web scraping API
The usual name for paying a service to fetch a page and hand back its text. That is exactly what this is. The account and the key belong to Okou rather than to you.
Scraping without writing a scraper
Nothing to build and nothing to maintain. You ask for the page in a message, and the fetching and the cleanup happen inside the run.
HTML to Markdown
The conversion people normally write themselves: a page of HTML in, clean Markdown out. It happens before the text ever reaches the agent, which is why a model spends its attention on the content instead of the markup.
AI or LLM web scraping
Scraping whose output is meant to be read by a model rather than parsed by code. It is the same fetch, judged on whether the text is readable rather than on whether a selector still matches.
No-code web scraping
Nobody writes a scraper here. You ask for the page in a message and the agent does the fetching, which is what people are looking for when they search for scraping without code.
Web scraping compared
Web scraping vs building your own scraper
Same output, a very different amount of work. On your own it is a service to choose or a fetcher to write, a key someone has to keep safe, and maintenance every time a page changes shape. Here you ask and read the answer, and access is a permission instead of a shared secret.
Web scraping vs the Firecrawl connector
Okou also ships a Firecrawl connector, and that is the right choice if you already have a Firecrawl account and want the usage there, or need something this service does not cover. The built-in one is right when you would rather just ask and get the page.
Web scraping vs driving a browser yourself
A browser you operate is more capable and much more work, and you still have to turn the page into readable text afterwards. Use the remote browser when the job genuinely needs clicking, and this when you just need what the page says.
The short version
The default way to read a web page on Okou. If the page is public and you want what is on it, this is one step, no account, and a charge per page. Anything that needs to sign in or click belongs to the remote browser.
Frequently asked questions
Do I need a Firecrawl account?
No. Reading a page is something any agent can already do, so you just ask. There is no account to create and no key to hold.
How is this different from the Firecrawl connector?
The connector works against your own Firecrawl account when you already have one and want the usage there. The built-in service is ready to use with nothing to configure.
Can it read pages behind a login?
No. It opens public pages with no sign-in. Use the remote browser, which can stay signed in and be taken over by a person part way through.
Can it read a whole website?
Not on its own. Each read is one page, so covering a site means asking page by page, and each page is charged. It is worth setting a limit.
What does reading a page cost?
Credits per successful read, with a higher charge for the stronger mode used on pages that resist a normal read. That mode is never chosen silently.
Is it safe to act on what comes back?
Treat it as information. Okou handles text from a web page as material to read rather than instructions to follow, which is what keeps a page from redirecting the agent that read it.
Do I still need a scraping tool?
For clean text from a page inside a workflow, no: you can ask an agent and get it. If you want a scraping vendor's own dashboard and full feature set, connect your own account through the connector instead.
How does the cost compare with a scraping plan of my own?
Scraping tools usually sell monthly plans with a minimum. Here reading a page consumes credits with no monthly commitment, which suits a workflow that reads a handful of pages a week and is worth measuring at very high volume.
Can it return Markdown for a model to read?
Yes. Clean Markdown is the default shape, which is what makes a page usable as context rather than as a wall of markup.
Do I have to write a scraper or maintain selectors?
No. There is nothing to write and nothing that breaks when a page is redesigned, because you are asking for the page's text rather than for the contents of a particular element.
How you ask for Web scraping
You describe what you want in plain language. The agent works out which service it needs and makes the calls.
“Read this page and give me the five things that matter, with the exact wording for each.”
“Pull the links off this documentation index, then read the three about pricing and compare them.”
“Check our competitor's pricing page every Monday and tell me only if something changed.”