Skip to content

Changelog

2026-11 - Product Matcher v2, picture matching, cost estimate, new Getting Started, new docs

  • Nov: Product Matcher v2 for every group. You start a match run yourself with Run match and stop it with Stop. Lines go in steps, Step 1 to Step 5. Each step has its own Floor and Max and a Rule, and you can add checks that run after a step matches. Results show Why matched, Decided by and the score of each Line. A line whose column a shop does not have is skipped for that shop and never blocks the run. See Set up a match, Steps, Floor and Max and Results and review.
  • Nov: Better picture matching. Photo check now compares the product pictures with a newer picture model, and a pair gets a text score, a photo score and a total score. Pictures cost EUR 1.50 per 1,000 and are counted only when photos are compared. See Photo check.
  • Nov: See the cost before a run. Next to Run match you see what the run can cost at most, what it usually costs, and each cost line (server, rows, pictures) under How is this charged?. The most it can cost is held in your wallet when the run starts, and what the run does not use comes back. If the wallet holds too little, the run does not start, nothing is charged, and the answer is a 402 that says how much is missing. See Run a match and Errors.
  • Nov: New Getting Started. Getting Started in the portal menu takes you from your first product address to your first rows, in a few steps. See Sign up and Your first data.
  • Nov: New docs. Every feature page now has three tabs: In the portal, Via API and Via MCP, and each API or MCP step names the key it needs, Read or Full. The MCP tools and Errors pages are built from the live service, the API reference lists only the routes a Read or Full key can call, and llms.txt lists every page for assistants.

2026-10 - Delivery status, daily matched file, sync uploads, review decisions

  • 23 Oct: Matching on a schedule. A group’s matching, classic or on the newer matching engine, can run by itself: after each scrape, daily, every N days or weekly at a set time. See Run matching on a schedule.
  • 22 Oct: Review decisions are kept. An approved pair stays matched and a rejected pair is not suggested again on every new matching run. After you decide pairs, Match quality is checked again by itself. Unmatch and Unmatch all reviewed… undo decisions. See Matched data.
  • 22 Oct: Held runs. On a group on the newer matching engine, a run whose result looks wrong in size is held, and your current results stay. See History.
  • 22 Oct: Find emails: 23 hours per run, then Continue. No daily count of websites any more. A run that stops at the time limit or at a daily guard is continued with Continue; nothing continues by itself. See When a run stops at a limit.
  • 22 Oct: Two key types. Keys are Full or Read. A Read key sees and calls only the read tools. See Tool catalog.
  • 21 Oct: Delivery status and webhook. Ask whether a group’s data for one day is complete, or get a signed webhook call. See Delivery status and webhook.
  • 21 Oct: Daily matched file. One fixed address per group gives the newest file of all matched rows. See Daily matched file.
  • 21 Oct: Stop, then delete. Deleting a scraper or a run whose run was stopped now ends the run first. A run that was not stopped answers 409 SCRAPER_RUN_ACTIVE. See Delete a scraper.
  • 21 Oct: Matching settings for the owner. On a group on the newer matching engine, the owner saves drafts of the text settings with a Full key or the portal. See Product matching.
  • 20 Oct: Sync uploads. Send an upload file as your whole catalogue with mode=sync: changed rows update, missing rows become inactive, matches and decisions stay. See Sync.
  • 20 Oct: Re-apply rules. Run a scraper’s current rules again on the rows of a finished run. MCP tools scrapewise_preview_reapply_rules and scrapewise_run_reapply_rules. See Tool catalog.
  • 18 Oct: Matching setup tools. Read and change the engine setup of a group on the newer matching engine over MCP. See Matching setups.

2026-10 - Matching pages: saved box spots, line filters, Match quality, Review fields

  • Match canvas: moved boxes keep their spot per group on every device; Reset layout puts them back. New: Only connected rows, legend items that show one type of line (Show all clears), and Lines you can add with an Unsaved · Undo state. The key line (★) cannot be deleted while its source has other lines. While a run is going, Stop sits next to Run. See How the portal draws your setup.
  • Matching tabs: History · Match the data · Results · Match quality. History, Results and Match quality appear after the first finished run. History is one table; Results has Matched / No match / Review with real counts and opens in the long view when the wide view cannot be built; Match quality shows only cards with a real number. See Matching in the portal.
  • Review: a Fields list puts chosen values under both pictures (Both sides first, then Left only, Right only, Technical), and Pairs narrows the review to some source pairs. Both are saved per group. See Review.

2026-10 - AI control: keep a group’s scrapers working

  • New: AI control for a group. On the schedule you choose (daily, twice a week or weekly) it checks the latest run of the group’s scrapers, finds failed runs, empty runs and missing values, tests a fix on live pages before saving it, and undoes a fix when the next real run is worse. In the portal: the group’s AI control dialog. Over REST: PUT /api/scraper/group/{id}/ai-control. See AI control.
  • It can repair scrapers that read the page with selectors (built automatically or by hand), and JSON-LD and API scrapers. AI-schema scrapers get an alert only for now. We are in a test phase: some problems are only recorded, not repaired, for now.
  • New: GET /api/scraper/group/{id}/ai-control/status shows the phase, the next check and the last result. It is also the new MCP tool scrapewise_get_scraper_group_ai_control_status (a Read key can call it).
  • Switching on needs your first wallet top-up. Switching off is always allowed.

2026-10 - Website contacts: emails from a business website

  • New: the websitecontacts endpoint. Send a list of business websites to POST /api/scraper/url-list-run (or the MCP tool scrapewise_run_scraper_url_list) with endpointTag: "websitecontacts" and get one row per website with the email addresses the business publishes on it. 2 credits per website read, charged only when an address was found or the business publishes none. Not part of the free requests. No portal screen yet. See Website contacts.

2026-10 - Fewer currency columns, shared per-unit columns

  • Changed: a CURRENCY_CONVERT rule now writes only its converted value. Each row gets one rate pair per target currency, for example eur_rate and eur_rate_date, shared by all conversions to that currency (scraper and group rules). New rows no longer get {outputField}_rate and {outputField}_rateDate. See Rate columns.
  • Changed: all conversions of one run use the same ECB rate table. A rule that only cleans a number in place (same from and to, overwritesRawField: true) writes no rate.
  • New row markers: RULE_RATE_CONFLICT:<to>_rate (two different rates needed on one row; both values kept, the pair left empty) and RULE_UNUSABLE:<to>_rate (the row already had its own rate column and it differs; that column is kept). See Row markers.
  • New: rules in different choices of one choice group may write the same output column, for example one unit_price for per 50 g, per 100 g and per piece. If no choice or several choices match, no output of the choice group is written.
  • New: ENUM_MAP can write a fixed value - an empty mapping with a default, for example unit = 50 g.
  • Existing rows are not changed: rows saved before this release keep their old column names, including the old _rate / _rateDate columns. Only some helper columns that start with __cc_ are no longer shown or returned anywhere (grid, REST, MCP, exports, samples). A group with old and new rows shows both the old and the new rate columns until the old rows are gone.

2026-09 - Upload a file

  • New: upload scrapers. Turn one CSV, JSON Lines or Excel file into a scraper whose rows become normal data in the group. In the portal: the group’s ⋮ menu → Upload. Over REST: POST /api/scraper/group/{groupId}/file-scraper and POST /api/scraper/{scraperId}/file. See Upload a file.
  • New MCP tools: scrapewise_create_file_scraper and scrapewise_update_file_scraper_file. Small files go inline; for a bigger file the tool returns one ready upload command. See MCP: Upload a file.
  • Limits: CSV and JSON Lines up to 200 MB, Excel up to 36 MB. Uploading is free.

Last updated 23 October 2026