AI Connections

AI Agents & MCP Overview

Connect Claude and other AI assistants to your Ultimate Web Scraper cloud workspace via MCP. Analyze, clean, and launch extractions in plain language.

Included with paid plans

AI connections are live for every cloud workspace on a paid plan — connect Claude, ChatGPT, or any MCP app today. See /pricing.

Ultimate Web Scraper speaks MCP (Model Context Protocol) — the open standard that lets AI assistants like Claude, ChatGPT, and Claude Code connect to your tools. Once connected, your AI scrapes with our cloud and works with the results conversationally. Here's what that actually looks like:

ClaudeUltimate Web Scraper
Scrape the entire catalogue of [your-store.com] — every product and variant.
Shopify Extractor412 products

Pulled the whole store — one row per variant.

1,180 rows · 7 columnsView table →
Real jobs the connection runs — scrape a store, map a site, collect leads, query a huge table.

Everything ends the same way: a clean table in your cloud workspace that you (and the agent) can query, clean, and export — no selectors, no code, no copy-pasting into the chat.

Start here

How it works

  1. One URL. Most apps need exactly one thing: the connector URL, pasted into their settings. Clients that don't sign in use an access token from the dashboard's Integrations page instead, with a paste-ready setup for each.
  2. Approve once. Sign in and pick the workspace and access level the app gets — or set the access level when you create the token. The connect guide has the per-app steps.
  3. Just ask. The agent lists your tables, queries and cleans them, checks on runs, starts new extractions, and answers product questions straight from these docs — all in conversation, all scoped to the one workspace you picked.

Which AI apps work?

Any MCP-compatible client — the connect guide has the exact clicks for each:

Cloud-only, by design

AI agents connect to your cloud workspace only. Data extracted locally with the Chrome extension stays on your machine — it is never exposed to agents unless you run the extraction in the cloud. This is the same privacy boundary as the rest of the product: local means local.

Selector-based scrapers (like List Extractor configurations) are still authored visually in the extension. Agents can run them in the cloud and use their results — building them stays point-and-click.

Requirements

  • A cloud workspace on a paid plan — see Credits & plans
  • Reading, querying, cleaning, and exporting data consume no credits
  • Mapping a site's URLs uses a small flat credit charge — quoted first, applied only when the scan actually finds URLs
  • Agent-started extractions consume credits like any other cloud run, always with an estimate first

On this page