MCP Scraper
Start for $1
Log inStart for $1

Now

November 2026

0.103.12

Choose which Reddit threads to fetch

Reddit Trends can show you the links first. Choose the discussions you want, then fetch their posts and comments together. Say “auto-approve” to search and fetch in one call.

Improvements
  • Preview candidate titles and URLs before opening threads, then fetch only the links you choose.
  • Keep related questions such as “Roofing Contractor?” in the results even if the title omits “recommendations.”
  • Use the date window as a discovery hint in review mode. Fetching verifies the post date or a captured comment date against that window.
Fixes
  • Recover candidates when the search engine returns an alternate results layout.
  • Check additional candidates when a dated search falls short.
MCP server
  • Set approvalMode to review in reddit_trending. The response contains the review ID and the proposed URLs.
  • Within one hour, call the tool with approvalMode set to approve, the original reviewId, and your approvedUrls. The original window and comment settings carry over.
  • Set approvalMode to auto to search and fetch immediately. Use reddit_thread when you already have one known URL.
  • Discovery, fetched threads, and returned comments are billed separately.
SDK
  • The Reddit Trends endpoint and Node, Python, and CLI bindings support review, approve, and auto modes.
  • Keep the review ID with the URLs you present in your app. Submit the selected URLs from that same review when the reader approves them.
Website 0.97.279–0.97.280

Browse product updates in Now

The website has a public home for releases and product launches. You can open and share an individual update from the feed.

Improvements
  • Use the Changelog link in the website footer to browse dated release notes.
  • Expand fixes, improvements, and developer notes inside an update.
Fixes
  • Match the website’s light palette and restore the space below the shared header.
0.103.11

Find Reddit threads by recent activity

Search across all time by default. When you select a date window, an older post can still qualify if a captured comment falls inside it.

Improvements
  • An undated search can return links without opening threads when comments are omitted.
Fixes
  • Verify dates on candidate threads when DuckDuckGo supplies the discovery results.
  • Keep a thread eligible when a captured comment falls inside the window, even if the original post is older.
MCP server
  • Dated searches verify activity even when the response omits comment text.
SDK
  • The public client uses the updated contract for undated links and verified post or comment activity.
0.103.7, 0.103.10

Recover sparse Reddit searches

Reddit discovery looks beyond the first result set when links are unresolved, the subreddit search is empty, or a 14-day search finds too few candidates.

Fixes
  • Keep organic Google results on Reddit for searches with a single explicit Reddit site scope. Retry a one-page light search with a narrower thread query if its results are all off-domain.
  • Look for replacement candidates when a 14-day result has no usable comments.
  • Refund the discovery charge when the run delivers no threads.
  • Retain captured results when the browser confirms the session has already closed.
SDK
  • The public client and desktop bundle picked up the 14-day search and comment delivery fixes.
0.103.8

Keep headings in website extracts

Website extracts preserve article headings and section links so you can trace details back to the original page.

Fixes
  • Retain substantive headers in the extracted body when they identify an article or section.
  • Preserve section links alongside the text they qualify.
SDK
  • The public client and desktop bundle picked up the extraction fix.
0.103.2 · Memory 2.0.5

Preserve curated Memory notes on re-upload

Re-uploading a file keeps the work you already did to organize the note.

Improvements
  • Keep titles, source attribution, tags, links, and capture dates on re-upload.
Fixes
  • Reject a concurrent overwrite when a newer edit already exists.
  • Keep local artifact paths out of Library captures. Deliberately captured short source text is still supported.
  • Resolve automatic note links only to existing notes in the allowed vault.
0.103.1

Collect more complete People Also Ask answers

People Also Ask continues past clipped or failed answers to collect the requested number of complete answers when the original search can supply them.

Improvements
  • Let acknowledged answers load while the tool opens subsequent questions.
  • Keep partial observations when collection stops short of the requested count.
Fixes
  • Retain late answers and their citations when another panel collapses or an interaction is interrupted.
  • Distinguish an empty result from an unreadable page or a consent failure.
  • Continue resolving available links while slower destinations finish.
0.103.0

Run multiple Transparent Commons publications

Publish to more than one subdomain from the same account and manage each publication independently.

Improvements
  • Editorial updates merge by default, so articles you leave out stay published. Choose replace mode when you intend to replace the whole edition.
  • Select a subdomain when reading from an account that owns multiple publications.
Fixes
  • Keep adjacent articles and their metadata when you update a single article.
  • Keep withdrawn content out of historical editions and reject stale writes or deletes.
MCP server
  • List your publications across pages, delete one article, or remove an entire publication.
  • Removed publication names stay reserved.

October 2026

0.102.4–0.102.6

Recover destination links in Google results

Google Search and People Also Ask can recover more destination URLs when Google wraps links in a redirect or a verification page.

Improvements
  • Apply link recovery to organic results, AI citations, PAA sources, and eligible secondary results, including video and local website links.
  • Keep original links for traceability and report unresolved destinations as partial.
Fixes
  • Preserve each result’s title, description, and snippet when two results share one recovered URL.
  • Handle Google consent before expanding PAA answers.
  • Keep link recovery cancellable and retain captured results if it reaches its deadline.
0.102.3

Keep crawl progress accurate under load

Background crawls recheck shared account capacity before resuming and keep their original job identity when they wait.

Fixes
  • Track progress as pages change and keep counts accurate across duplicate writes, replacements, and lost acknowledgments.
  • Include preparation and database waits in crawl deadlines while reserving time to publish saved results.
0.102.0–0.102.1

Use one hosted MCP protocol

Hosted tool calls use one protocol for crawls, Maps, and other operations. Lead enrichment jobs can now be cancelled through a dedicated tool.

MCP server
  • Direct hosted clients must use protocol 2026-07-28. Update and restart public clients before connecting to the strict endpoint.
  • Use an operation’s dedicated status tool with its existing job or run ID. The earlier MCP Task handles were removed.
  • Use lead_list_enrich_cancel to cancel a lead job while retaining its progress and billing state.
SDK
  • SDK 0.58.1 updates the Node, Python, and CLI clients and corrects direct HTTP examples to include the required metadata envelope.
0.101.0–0.101.2

File private bug reports and export GA4 data

Prepare a reproducible bug report from a failed tool call, or export Analytics reports across multiple properties.

Improvements
  • Review a report with the failed call, error, client details, and reproduction steps before submitting it. Submitted reports include a receipt and redact credentials.
  • Export GA4 reports across selected or discovered properties with pagination, coverage details, and explicit per-property failures.
  • Discover Analytics accounts, properties, and streams. Access administration requires separately verified Google permission.
Fixes
  • Preserve time zone, pagination, sampling, thresholding, and quota metadata in ordinary GA4 reports.
  • Fix an HTTP 400 that could occur after successful discovery in npm, npx, and desktop clients.
SDK
  • Public client release history and desktop downloads became accessible without private repository access.
0.100.1–0.100.2

Keep more Google Maps review and listing details

Google Maps tools preserve short reviews, line breaks, categories, and addresses across more listing layouts.

Fixes
  • Recognize both Places and Businesses results and read categories without an industry keyword list.
  • Keep categories and addresses for businesses with no ratings.
  • Preserve collected reviews when a browser closes and respect the requested review limit across resumed runs.
  • Resume eligible partial place results without reacquiring completed runs.
0.100.0

Open known Reddit threads and read saved crawl pages earlier

A Reddit URL can go directly to thread collection. Site crawl pages become readable as they are saved, while the ZIP is still being assembled.

Improvements
  • Use a known Reddit thread URL without running a discovery search first.
  • Read a complete single-page extract behind its preview and download its JSON on demand.
  • Read large site and archive text results in consecutive windows.
Fixes
  • Capture replies behind supported nested controls without counting a reply twice.
  • Return usable partial Reddit comments with coverage details when collection comes up short.
  • Show a single completed extraction in history while keeping separate requests to the same URL distinct.
0.99.0

Open retained search exports without running the search again

Completed searches return full JSON before optional exports. Use the returned delivery information to retrieve a saved export.

Improvements
  • Find available exports in delivery.availableExports and request a Markdown download when needed.
  • Keep completed Maps results and transcripts available if an optional delivery or history step fails.
MCP server
  • Breaking change for Google Search: use one page with a date filter or saved search identity. To request two pages, remove both and expect organic results.
SDK
  • Read or download a returned export handle without repeating acquisition.
0.97.2

Verify Google Maps business identity across profile layouts

Maps can verify a business using a structured profile ID when the listing omits a direct Maps link.

Fixes
  • Reject conflicting business identities before collecting or merging profile data.
  • Recover detailed reviews on supported layouts without a direct link.
  • Mark the result partial if the full set of requested material is unavailable.
0.98.0 · Memory 2.0.3

Move the public SDK to hosted execution

Public clients send authenticated requests to MCP Scraper’s hosted service. The CLI keeps its installation helpers and local result downloads.

SDK
  • Breaking change for library callers: migrate local scraping and generation calls to public MCP results and delivery handles.
  • The CLI’s local API command acts as an authenticated relay to the hosted service.
  • The public Memory compatibility bridge continues to use existing authenticated configurations.
0.97.0 · Memory 2.0.0

Move from scheduled actions to on-demand work

Recurring actions, workflow schedules, assistant reminders, and automatic vault cleanup were removed. Historical notes, results, and billing records stay available.

Improvements
  • Use the retained tools for on-demand workflows and continue reading or downloading saved results.
MCP server
  • Scheduled-action and scheduled-run tools were removed from the hosted and standalone Memory catalogs. Saved artifact templates remain available through their tools.
SDK
  • Scheduling methods were removed. Find template methods in the artifacts namespace and connected-service routes under /connections.
0.96.0

Retrieve complete results behind short previews

Large results can stay readable in chat while preserving the complete source material for download or a follow-up read.

Improvements
  • Read full retained results through artifact_describe, artifact_read, or artifact_download.
  • Continue text reads from the returned offset and download binary files in their original format.
  • Keep full transcript text alongside structured results and readable previews.
Fixes
  • Keep source titles, descriptions, citations, and snippets alongside captured People Also Ask answers.
  • Preserve Maps review text across fallback and saved retries after verifying the business identity.
SDK
  • For normalized SERP captures, read harvestResult. Raw providerPayload is no longer part of the public result.

September 2026

0.95.2

Copy the setup format for your MCP client

Dashboard setup instructions give each client a labeled command or configuration format.

Improvements
  • Choose the Codex command or TOML configuration and see which destination it belongs in.
  • Get accessible feedback when you copy the setup snippet.
Fixes
  • Report missing directory jobs as not found and remove retry advice when the underlying error cannot be retried.
0.94.0

Resume saved Google Maps place runs

A timed-out client can reopen the same Maps place run, inspect saved results, and explicitly resume eligible work.

Improvements
  • Page through saved reviews and images instead of reopening a browser to read completed material.
  • Choose which place material to request, including reviews, configured services, service areas, and photos.
  • Resume image ZIP packaging from saved image URLs.
MCP server
  • Use the saved run status and resume tools after a client timeout. Saved output remains scoped to its owner.
0.92.0–0.93.0

Choose light or full Google Search output

Light mode returns organic listings. Full mode also reports available search features from the captured page.

Improvements
  • Read organic positions, URLs, titles, and descriptions in light mode.
  • Check feature status in full mode to see which modules were available.
  • One page is the default. Request the second page explicitly when needed and check the delivered coverage.
0.91.0

Transcribe a public TikTok video

Turn a public TikTok video or short share URL into a text transcript with timing and source details.

MCP server
  • Use tiktok_video_transcribe to retrieve full text, timed chunks, and a speech-quality signal.
Improvements
  • Check the source details and returned quality signal before quoting a transcript.
0.89.0

Inventory public Facebook Reels

Collect Reel URLs from a Facebook profile or Page for a later review or transcription batch.

Improvements
  • Deduplicate results against URLs you already collected.
  • Keep available URLs when a scan ends early and use the stopping reason to understand the coverage.
Fixes
  • Dismiss optional login dialogs that obscure public Reel results.
0.87.0

Request a second search page with People Also Ask

Collect a second organic page alongside questions and answers from the original first page.

MCP server
  • Set pages to 2 in a synchronous or background PAA harvest when you need more organic results.
Fixes
  • Keep first-page material when the optional second page fails to load.
SDK
  • Read requested and captured page counts to distinguish complete from partial pagination.
0.86.0

Open dedicated Research and CRM workspaces

Choose a Research or CRM vault and work with the records that fit its purpose.

Improvements
  • Manage people, organizations, activities, tasks, deals, and pipelines in CRM.
  • Review duplicates during imports and retain stable IDs when handing work off.
  • Open legacy Analytics CRM bookmarks in CRM Settings with their recognized X-Ray context.
MCP server
  • Search CRM work and list or update pipelines through governed Memory tools.
0.83.0

Publish X-Ray measurement rules and run approved experiments

The installed X-Ray Pixel can receive new event rules without a site code change and supports separately approved presentation experiments.

Improvements
  • Publish versioned measurement rules and inspect selector health on dynamic pages.
  • Use canary and production experiment states and record observed exposure.
  • Approve the exact experiment revision before activation. Publishing measurement rules alone leaves the experiment inactive.

August 2026

0.68.0

Trace conversions back to consented first-party activity

X-Ray connects consented page activity, calls, and CRM events into a journey you can inspect.

Improvements
  • Review first, last, and ordered attribution touches alongside engagement measurements.
  • Define page, click, phone-link, and same-origin form events for existing Pixel installs.
  • Receive supported phone and CRM events through signed webhooks.
Fixes
  • Initialize the Pixel when a site inserts it after hydration.
  • Return the original conversion ID when the same conversion is submitted again.
0.64.0

Enrich a lead list you already have

Inspect a CSV, TSV, or Excel lead list before starting enrichment, then retrieve a private export from the completed run.

Improvements
  • Check the import before committing to a paid run, and retain row-level progress if work is interrupted.
  • Use the source information on each returned contact or company match to review the evidence.
  • Opt in to exact-domain email discovery or owner and leadership discovery when the work needs it.
MCP server
  • Use lead_list_enrich for your supplied list. CSV and Excel results from this release are retained for seven days.
0.57.0

Keep Transparent Commons images available

Transparent Commons can host its own article images so published pages keep displaying them when the source host changes the original link.

MCP server
  • Use commons_host_image with a public HTTPS URL or supported image bytes to obtain a Commons image URL.
Fixes
  • Match search terms as words and improve relevance for short queries.
  • Resolve See Also links when an article’s slug differs from its title.
0.54.0

Generate long-form articles and personal authority pages

Two artifact templates add a long-form editorial layout and a personal authority page with configurable branding.

Improvements
  • Give a long-form article a media rail, expandable disclosures and FAQs, and a table of contents that follows the reading position.
  • Share an article or copy its citation from the rendered page.
  • Choose a personal page layout and apply a supported custom brand color.
0.52.0

Keep website images and Google Maps listing photos

Opt in to media collection when a website or Maps job needs images alongside text and structured fields.

Improvements
  • Collect lazy-loaded and responsive website images, collapse duplicate variants, and inspect the ZIP manifest.
  • Request owner photos or all available photos for a Maps listing and review the evidence for ownership.
  • Separate logos from partner marks, awards, certifications, and other visible trust evidence.
Fixes
  • Exclude challenge pages and non-page responses from website text extraction.
0.50.0

Compare page engagement in X-Ray reports

Content reports include visible time on page, scroll depth, and the coverage of measured pageviews.

Improvements
  • Compare time on page, bounce, exit, and engagement across paths.
  • Search paths and sort across the complete report before opening or exporting a page of rows.
Fixes
  • Include report rows in Markdown exports and keep engagement measurement separate from interaction counts.
0.44.0

Submit Local Sourcebook listings and propose Commons entities

Local Sourcebook gives subscribers a way to prepare and revise business listings. Transparent Commons adds searchable public entities with source evidence.

Improvements
  • Inspect acquisition coverage, public images, and captured reviews behind a Local Sourcebook listing.
  • Keep the last approved listing public while a refresh is being acquired or reviewed.
MCP server
  • Prepare and validate a listing draft before submission.
  • Search Commons entities, read an existing article, and propose a new or revised entity with supporting sources.

July 2026

0.38.0

Read individual files from a site export

Open a site crawl ZIP to see what it contains and read a selected text file from the export.

MCP server
  • Pass the bundleUrl from check_site_export to archive_read. Omit path to list the files.
  • Select a UTF-8 path to read the file in windows or explicitly save the selected material to Library.
0.34.0

Discover Reddit threads by topic

Reddit Trends launched with topic and subreddit searches and the option to collect comments on selected results.

Improvements
  • Discover candidate discussions through a Google search scoped to Reddit.
  • Use a recency window to focus on discussions from a recent week or month. The date defaults changed in later releases.
MCP server
  • Discover discussions by topic or subreddit with reddit_trending.
0.27.0

Read localized Google Maps profiles and services

Google Maps discovery walks local listings and their profiles for the requested location.

Improvements
  • Read profiles in ranked result order.
  • Request configured services and areas served when the listing provides them.
0.11.0

Connect your services to MCP Scraper

The connected-service catalog added Gmail, Google Drive and Calendar, GitHub, Slack, and other integrations.

Improvements
  • See the available read and action tools for each connected account on the integrations screen.
  • Keep connected-account history in dedicated vaults.
MCP server
  • Use read_service_connection for approved read operations exposed by a connected account.
  • Live Gmail, Slack, Calendar, and Zoom actions require the account owner to enable write access explicitly.