AI Connections
MCP Tool Reference
Every MCP tool a connected AI agent can use in Ultimate Web Scraper — querying tables, cleanup, exports, automations, site discovery, and Shopify extraction.
A connected AI agent works through a set of named tools — you ask in plain language, the agent picks the right one, and most clients show each call as it happens. You never call these yourself, but knowing what's on the shelf helps you ask for the right things — and Prompts to try turns each of them into a ready-made ask.
The scraping toolkit at a glance
What a full-access agent can scrape, with the tool it calls to do it:
Similar pages
scrape_pagesOne row per URL from similar pages of one site — products, listings, articles, job and real-estate pages. Structured data captured automatically, no selectors; page metadata and phone numbers can join as extra columns.
Shopify stores
extract_shopify_storeAn entire Shopify store — or just the collections you name — as one row per variant: price, SKU, options, images, availability.
WooCommerce stores
extract_catalogAn entire WooCommerce store — or the categories you name — as one row per variant: price, SKU, stock, images, variations.
Contacts & leads
extract_contactsLead lists across many different sites — one row per site with emails, social profile links, or both in one run. A deep scan follows internal links to reach contact and team pages.
Map places
extract_map_placesName, rating, review count, address, phone, website and opening hours from a list of place pages — typically gathered with the extension's List Extractor. One row per place.
Site discovery
discover_sitemap · select_sitemap_urlsMap every page a site publishes, grouped into sections, then pick which ones to scrape — the picked URLs feed straight into scrape_pages. No URL list needed.
Website analyzer
analyze_websiteCheck any site before scraping it — detects the platform, finds the sitemap, and picks the approach that works.
Automations
run_automation · update_automationRun any saved automation — cost estimated from its run history first — check its schedule and latest results, rename, reschedule, pause or delete it. A deleted automation keeps its tables.
Data cleanup
deduplicate_rows · delete_rows · merge_columnsTidy the table after the scrape — rename or merge columns, drop the ones you don't need, delete unwanted rows and remove duplicates.
On top of the ready-made extractors, the agent also:
- runs any point-and-click scraper you've built in the extension and sent to the cloud
- keeps any job on a recurring schedule
- answers "how do I…?" questions straight from these docs
Reading & analysis — every connection
The tools below are available to every connection, including read-only ones.
| Tool | What it does |
|---|---|
get_workspace_info | Your plan, credit balance, and the connection's access level |
list_tables | Lists your result tables, grouped by the automation that produced them |
get_table | One table's columns, row count, and a small sample of rows |
query_table | Filters, aggregates, groups, and sorts a table of any size — the full table never enters the conversation |
export_table | Generates a download link in the format you pick — CSV, JSON, or Excel. Tables produced by a Shopify extraction can also export as a re-import-ready Shopify product CSV |
list_automations | Your automations, their schedules, and their status |
get_automation | One automation's setup at a glance — what it scrapes, its schedule state, recent runs, and the latest results table |
list_runs | The run history of an automation, optionally limited to runs since a given time — "what ran this week?" |
get_run | A run's live progress and row count so far, plus links to watch it or open the results table. On a failed run it adds what went wrong, how many rows were kept, and what to try next |
search_docs | Searches this documentation to answer product questions |
get_docs_page | Reads a single docs page in full |
list_docs | The documentation's table of contents |
Cleaning & operating — full access
| Tool | What it does |
|---|---|
run_automation | Runs an automation — it estimates the cost from the automation's run history first, and starts only after you confirm |
update_automation | Renames an automation or reworks its schedule — sets or clears the cron, pauses it without losing the schedule, resumes it — and reports back the resolved next run |
delete_automation | Deletes an automation and its schedule, with confirmation first — its data tables survive |
stop_run | Stops a run in progress |
export_table_to_google_sheets | Sends a whole table to a new spreadsheet in your Google Drive (uses the Google Sheets integration) |
update_column | Renames a column or changes its type |
delete_columns | Deletes columns you don't need |
merge_columns | Merges 2–5 columns into one (street + city + zip → Address) |
delete_rows | Permanently deletes the rows matching a filter — it reports the match count first and deletes only after you confirm |
deduplicate_rows | Removes duplicate rows, keeping the first occurrence — with the same count-first confirmation |
Extracting — full access
| Tool | What it does |
|---|---|
analyze_website | Free first look at any site you want scraped: identifies the platform, checks for a sitemap, and returns the extraction strategies that apply |
scrape_pages | Scrapes similar pages of one site — products, listings, articles, job posts, real-estate pages — into one row per URL, the structured data captured automatically, no selectors. Page metadata and phone numbers can join as extra columns. Takes up to 500 pasted URLs, up to 4,000 from a table's URL column (with an optional row filter), or a sitemap discovery. The credit estimate comes first — with a free peek at the columns the first page yields, when that page is readable — and the scrape starts only after you confirm |
extract_contacts | Builds a lead list across many different sites — one row per site with emails, social profile links, or both in one run, one column per platform. An optional deep scan follows each site's internal links one level to reach contact and team pages. Up to 500 pasted URLs or 10,000 from a table column; credit estimate first, runs on your confirmation |
extract_map_places | Turns map place-page links — typically gathered with the extension's List Extractor — into one row per place: name, rating, review count, address, phone, website, and hours. Search and results links are rejected up front with guidance. Up to 500 pasted URLs or 4,000 from a table column; credit estimate first, runs on your confirmation |
create_extraction | Legacy starter kept so saved workflows keep working — new extractions go through the three tools above |
discover_sitemap | Maps a site's URLs into a grouped overview — the starting point for "scrape this whole site" or "just the winter collection". A fresh scan quotes a flat 25 credits and runs once you confirm, charged only when it actually finds URLs; a recent scan of the same site is reused free, no confirmation needed. The same idea as the extension's Sitemap Explorer, run for you by the agent |
select_sitemap_urls | Filters a discovery — by group, search term, or exclusions — into a URL table that feeds straight into scrape_pages |
list_shopify_collections | Lists or searches a Shopify store's collections by name |
extract_shopify_store | Extracts a Shopify store's catalogue — the whole store or chosen collections — as one clean row per product variant. Estimates credits and asks before running |
list_catalog_collections | Lists or searches collections (Shopify) or product categories (WooCommerce, Magento) on a store with a structured catalogue. On WooCommerce and Magento this runs only after you pick specific categories — analyze skips the list so it stays fast |
extract_catalog | Extracts a Shopify, WooCommerce, or Magento catalogue as structured product data. All three use the same scope as the UI: whole catalogue (default), or chosen collections/categories. preview:true samples the chosen scope for free |
What each extraction type captures, and its URL limits, live in AI-started extractions.
Ready-made prompts
The connection also includes starter prompts — one-click workflows like "discover and scrape a site", "scrape a Shopify store", or "clean up a table" — that appear in Claude's connector menu and as slash commands in Claude Code. Only ones your access level can run are shown.
Credit & safety behavior
- Reading, querying, cleaning, and exporting consume no credits. A fresh site discovery uses a flat 25-credit charge, only when it actually finds URLs — a recent scan of the same site is reused free.
- Anything that spends credits or removes data happens in two steps: starting an extraction, running an automation, or a fresh site discovery shows you a credit estimate first — and deleting rows or an automation shows a match count or a confirmation — and nothing runs until you confirm. An extraction never begins silently.
- Oversized results are never dumped into the chat — the agent is asked to narrow its query instead.
More in the security model.
Related
What AI Agents Can Do
Connected AI agents can query scraped tables of any size, clean columns, export data, run and schedule automations, and answer product questions from the docs.
AI-Started Extractions
Ask Claude or any MCP agent to scrape product pages, build lead lists with emails and social links, or extract map place data — with credit estimates before anything runs.