Yelp Business Scraper — Source Fields, Website Signals & MCP

Submit direct Yelp business URLs for business research and CRM preparation. Collect available name, ratings, review counts, categories, price range, address, phone and hours; optionally inspect a listing-returned business website for public contact, technology and SEO signals. Fields depend on the source response and can be unavailable.

Free:$1.50 per1,000 useful business rows; Bronze–Diamond:$1 per1,000, plus startup and separate platform/proxy resources. Startup is$0.00005 perGB of run memory, minimum1 event; at256MB it is$0.00005. See the Pricing tab for the current tier. A row may contain only the basic fields the source returned; this is not a guaranteed full-contact lead.

First run: one business, basic fields

{"businessUrls":["https://www.yelp.com/biz/tartine-bakery-san-francisco"],"maxResults":1,"extractWebsite":false}

Read the Dataset and SUMMARY together. Check businessName, rating_normalized, reviewsCount_int, sourceMode, coreVerification and collectedAt. Do not equate a successful run with complete fields or independently confirmed company identity.

Only direct /biz/... URLs on supported Yelp domains are accepted. They are canonicalized and deduplicated before source access. The runtime default is10 unique inputs, with a maximum10000. Empty input succeeds with zero result events; invalid-only input or a run with no substantive business extraction fails with free diagnostics. A small run-start charge still applies.

Sources and limitations

The production path currently uses structured extraction of the requested business page. sourceMode:structured_extraction distinguishes it from an optional configured structured API source (structured_source). coreVerification:source_returned_fields_not_independent_identity_check means the fields were returned for the submitted URL; their identity was not separately established from original Yelp HTML.

At the October2026 checks, direct Yelp HTTP and a residential-proxy probe returned403/challenge pages. The working extraction path does not require a customer-supplied external key, but depends on an external source's availability and field coverage. No guarantee of native HTML access, all pages, every country or exact freshness is made. Source-returned values can require manual verification. Names, cities or guessed websites derived solely from the URL slug do not become paid Yelp records.

Keyword/location search is not configured in production and is not offered in the form. Legacy searchTerms calls receive SOURCE_ERRORS; a search-only request fails with zero business result events. Search results are not substituted with URL guesses. This Actor does not promise whole-city discovery, full reviews, review-history monitoring or an indexed Yelp database.

Input and field coverage

Option	Default	Behavior
businessUrls	sample URL in form	At most10000 direct business URLs.
maxResults	10	Cap unique inputs analyzed and rows delivered; remaining URLs go to continuation.
maxConcurrency	2	Bounded business batches,1–5. Event budget can reduce the batch size.
timeout	90	Total processing budget per business,10–300seconds. Run deadline reserves report time.
maxRetries	1	0–5 retries within the total source deadline. More retries can increase time and resource cost.
extractWebsite	true	Analyze an available listing-returned website.
extractContactEnrichment	true	Public email/phone/social/SEO/schema signals from a fetched website.
extractHoursIntel	true	Normalize available schedule text. Missing schedules stay unknown.
extractAmenities	true	Pattern-based amenity observations; verify ambiguous or negated source text.
extractLeadScore	true	Optional prospecting heuristic and reasons.
extractOutreachPitch	true	Draft text only; no message is sent.
guessEmailPatterns	false	Explicitly separated speculative addresses; never verified email contacts.
websiteDiscoveryFallback	false	Opt-in search candidate, labeled unverified association.
extractAge	false	Available Wayback snapshot observation, not business founding date.
extractDomainAge	false	Optional public certificate-history signal, not confirmed domain registration age.
extractGeocoding	false	Approximate address coordinates when enough source data is available.
excludeChains	false	Filter records scoring at least50 on the chain heuristic; skipped rows are not billed.
exportFormat	default	Full JSON or flattened CSV-friendly records.
writeSummary	true	Controls optional TOP_LEADS; operational counts/diagnostics are always saved.

Business fields include available businessName, rating, rating_normalized, reviewsCount, reviewsCount_int, categories, priceRange, address, phone, website, hours and image. Rating is0–5; invalid/non-finite normalized values are null. Missing website/address/contacts remain unknown, not proof that the business lacks them.

Website findings include available websiteCheckStatus, websiteFinalUrl, websiteBodyTruncated, website_tech_stack, emails_from_website, phones_from_website, social_profiles and SEO/schema fields. A403, challenge or non-HTML response does not establish a healthy website or paid technology finding. Technology matches are passive pattern observations, not executed scripts or independent backend proof. websiteAssociation distinguishes a listing-returned link, an unverified search candidate and unavailable association.

Exact is_open_now stays null without a verified timezone and exception-hours context. A state-derived zone is exposed only as timezoneEstimate. Weekly hours can exclude holiday changes or overlapping intervals; verify the original schedule before operational use. waybackClosestSnapshotDate and waybackSnapshotAgeYears describe a returned snapshot closest to the requested historical date, not the earliest listing or the age of the business.

Popularity, presence, chain and lead scores are heuristics derived from available fields. leadScoreMethod:heuristic, reasons and audit-method fields expose this limitation. A chain filter can misclassify a business. Review and personalize any generated outreach draft. Guessed emails are kept separately from extracted addresses and are not selected as verified email contacts.

Results, billing and continuation

Only successful source-returned rows with a business name and substantive rating/review evidence, phone or address enter the paid Dataset. There is one apify-default-dataset-item event per useful row; the event title is business-scraped. Do not manually charge the synthetic event again. Duplicate URLs, invalid inputs, failed extraction, URL-only guesses, skipped chains, summaries and diagnostics incur no result event.

Free key-value records:

• SUMMARY: requested/validated/delivered/duplicate/failed/pending/skipped counts, status, coverageComplete, limits and nextInput.
• FAILED_BUSINESSES: rejected or insufficient-source inputs with safe reason codes.
• SOURCE_ERRORS: unsupported legacy search diagnostics.
• SKIPPED_BUSINESSES: heuristic filter exclusions.
• RETRY_INPUT: source-failure candidates. Run only if businessUrls is non-empty, after reviewing access and limits; not every source failure is temporary.
• TOP_LEADS: optional useful rows sorted by the heuristic score, maximum20; not a guarantee of contactability or commercial intent.

status distinguishes complete, partial, empty and source_unavailable. A budget/time/result cap leaves unprocessed URLs in SUMMARY.nextInput. Use that input for a subsequent run after adjusting the limit. If a failed extraction also needs another attempt, review RETRY_INPUT. Budget limits are checked before source work; separate platform/proxy usage can still apply.

REST API, MCP and agent contract

Use this Actor when you already have Yelp business URLs. Do not ask it for unconfigured keyword discovery, all city businesses, complete reviews, ownership verification or verified private contact details.

Start with POST https://api.apify.com/v2/acts/apivault_labs~yelp-business-scraper/runs, authenticate using Authorization: Bearer <APIFY_TOKEN>, and send the JSON input above. Poll GET /v2/actor-runs/{runId} to a terminal status. Read GET /v2/datasets/{defaultDatasetId}/items and key-value records at /v2/key-value-stores/{defaultKeyValueStoreId}/records/{recordKey}. Read diagnostics even if the run failed. Dataset items support ?format=csv&clean=true and ?format=xlsx&clean=true; credentials belong in your client configuration.

Connect an authenticated MCP client to:

https://mcp.apify.com/?tools=apivault_labs/yelp-business-scraper

Discover tools/list, then call apivault_labs--yelp-business-scraper with the same inputs. A short waitSecs may return before completion; poll the returned run instead of creating duplicates. Use discovered get-dataset-items and get-key-value-store-record to read the Dataset, SUMMARY and FAILED_BUSINESSES.

Agents should begin with an explicit small cap, set extractWebsite:false when only listing metrics are needed, and keep email guessing/search-candidate discovery off unless explicitly wanted. Report missing fields as unknown, read provenance and website status, and include partial scope/continuation in the answer. Do not describe source-returned extraction as independently verified native Yelp HTML. Only the useful Dataset rows incur result events; a source failure or an empty Dataset is not evidence that the requested business does not exist.