#Browsing the web

Arachne opens real websites and operates them the way you would — following links, searching within a site, signing in to services you connect, and pushing through ordinary web friction to reach your goal.

#Overview

Arachne is Nourva's hands on the open web. When a task needs a real browser — pages that have to be opened, links that have to be followed, content that has to be read in place — Arachne does it for you. You describe the outcome you want in plain language, and Arachne works through real websites and applications to get there, taking in what is on the page, carrying out each step, and confirming the result before continuing.

This is what separates browsing from a simple answer. An answer tells you something; browsing goes and does something on a live site. Arachne takes in a page much as you do, handles the small obstacles that slow people down, and reports back with a verified result. You never script the clicks or name the steps — you name the destination, and Arachne finds the path.

#What you can do

On a live page, Arachne can:

  • Open and move through real websites — go to a site, follow links, and travel from page to page toward your goal.
  • Search within a site — use a site's own search to find the specific page, listing, record, or item you need.
  • Read content in place — take in what a page actually says, including details buried deep in long pages, and act on what it finds.
  • Sign in to services you connect — access services you have authorized so it can work on pages that require you to be signed in.
  • Operate pages like a person — click, select, scroll, expand, and interact with the controls on a page to make it do what you want.
  • Handle ordinary web friction — work through the small obstacles of everyday browsing so you do not have to babysit each one.
  • Gather and compare — collect information across several pages or sites and bring it together into a result you can use.
  • Report back with something you can see — present what it found or did in whichever of the eight formats fits: table, chart, report, comparison, dashboard, media gallery, alert, or source pack.

Everything Arachne does on a page is automatically checked before it is reported done. If a step does not actually succeed, Arachne notices, says so honestly, and adjusts — so a page it reports as "reached" was really reached, and information it reports as "found" was really there.

#How to use it

  1. Describe the outcome, not the clicks. Tell Arachne where you want to end up or what you want to come back with — the page to reach, the information to gather, the thing to do. Let it work out the navigation.
  2. Give what you already have. If you know the starting site or the exact link, include it. A good starting point saves time and keeps Arachne aimed at the right place.
  3. Connect the services you need. For pages that require you to be signed in, authorize the service first so Arachne can reach what is behind it. It only signs in to services you connect.
  4. Choose a permission level for the run. Pick how much freedom Arachne has: Look only to read and inspect without changing anything, Plan first to approve an approach before it acts, Ask before acting to confirm each change, Work in your workspace to act freely inside your project, or Act automatically to carry the whole task through. Start cautious on an unfamiliar site; give more freedom once you trust the run.
  5. Watch the first run. You can watch Arachne work live — see which pages it opens and what it does on each one — so you can steer if it heads somewhere you did not mean.
  6. Stop anytime. Stopping always works immediately, at any permission level. If a run drifts, stop it, clarify the goal, and start again.
  7. Pick up the result. When Arachne finishes, it presents what it found or did. Everything it produces is saved as an artifact you can reopen, reuse, and export.
Tip

When you want to be sure nothing on a site changes — for example while Arachne explores or reads — set the run to Look only. Arachne will read and inspect but never alter anything on the page.

#Examples

What you ask:

text
Go to my supplier's site, find the current price and lead time for part number 4471, and bring it back.

What you see: Arachne opens the supplier's site, uses its search to locate part 4471, reads the listing, and returns a small table with the current price and lead time — with the source page noted so you can check it yourself.

What you ask:

text
Search this site for every job opening in their design team and list them out for me.

What you see: Arachne searches within the site, follows each result, reads the postings, and returns a table of the openings — title, location, and the link to each — gathered from across the site's pages.

What you ask:

text
Sign in to my account on the service I connected and check whether my latest invoice has been paid.

What you see: Arachne signs in to the service you authorized, navigates to your billing area, reads the latest invoice, and reports its status back as a short, clear notice — pausing for your permission first if your permission level asks it to.

#Tips

  • Be specific about the destination, not the route. "Find the warranty terms for this product" gets you further than a list of pages to click. Arachne is good at finding the path; it just needs a clear target.
  • Give the link if you have it. Starting Arachne at the exact page, instead of the homepage, removes guesswork and gets you to the result faster.
  • Match the permission level to the risk. Reading and gathering are safe at Look only. Reserve the freer levels for sites and tasks you already trust, and use Ask before acting whenever a page involves submitting or changing something.
  • Watch the first run on a new site, then loosen up. Once you have seen Arachne handle a site correctly, you can give later runs more freedom with confidence.