About Pluck / Built by Kris

Small tool.
Less friction.
More flow.

For the in-between work: the lookup, the tab switch, the detail you need to carry from one place to another.

Pluck’s purple pick-and-drop icon

A little bridge between your tabs.

Pluck is a Chrome extension by Kris. It brings a repeatable workflow to a familiar responsibility: checking information on another website before updating the page in front of you.

Configure the source page, lookup, and extraction once. With Pro Auto mode enabled, Pluck recognizes a matching page and retrieves the related information from a website or API. Fresh results appear beside your current record, ready to compare.

Review the differences, fill mapped fields without retyping, and save in your own application. You can also drag individual values or run a batch of references into a CSV.

More from Built by Kris
THE IDEA

Build once. Use again.

A saved profile holds your source, lookup steps, extraction rules, and optional target fields. Give a repetitive task a repeatable route.

THE EXPERIENCE

Keep the work in view.

A Chrome side panel keeps results beside your page. Cards show what was found and, when mapped, what matches or differs.

THE CHOICE

Add only what you need.

Start with selectors or a webhook. Add automatic triggers, transformations, or optional AI when your workflow calls for them.

A clear view of the product

Made to be configured.
Made to be understood.

Pluck does not come with a promise that every website works out of the box. Its picker, tests, fallback selectors, and run logs help you build and inspect your own workflows.

The examples on this site use fictional data. Named AI providers describe supported connections; they do not imply endorsement or partnership.

This showcase describes Pluck version 0.2.0 with Free and Pro plans. Pro starts at $9 USD per month for two devices, with checkout through Lemon Squeezy. Pluck is available on the Chrome Web Store.

A little less busywork

Configure. Open.
A little magic. Save.

Pluck is now available on the Chrome Web Store. Install it, configure your first lookup, and give your everyday work a little less busywork.

Chrome 121 or later · Free to get started

Add Pluck to Chrome
Ask AI
The little things keep getting better

Pluck release notes

Pluck: Scrape, Lookup & Drag-to-Fill. A Chrome (Manifest V3) extension that looks up values on a website, API or AI model and returns them in a side panel to drag into, or autofill on, another page.

Newest first in the summary table; each version below lists its changes, grouped by area. Dates reflect the original commit or development dates. Versions 4.2.0 and 4.1.0 are planned. Version 4.0.0 is in progress. Versions 3.0.0, 3.0.1 and 3.1.0 are now live on the Chrome Web Store.

Version index · newest first
VersionDateTheme
4.2.02026-10-08Planned · Retry only what failed, re-run failed rows, ready-made patterns, row-number condition, end-of-day reminder
4.1.02026-10-08Planned · Flow control: run-if conditions, set a value, repeat while, continue in the same tab
4.0.02026-10-08In progress · Bulk scraping: loop only some steps, wait-and-retry, per-site stop, running scrape table
3.1.02026-10-05Live · Clone lookups, rename without breaking references, add and join API values
3.0.12026-10-05Live · Fixes for multi-site Autofill
3.0.02026-10-04Live · Page scanner and Mapper, filling several sites at once
2.2.02026-10-02Export/import single lookups, Retry handles captchas
2.1.02026-10-02Outbound alerts, printable verification reports
2.0.12026-10-02Debug report and vault-protected AI keys
2.0.02026-10-02Multiple website lookups, scheduling, checks, admin policy
1.5.02026-10-01Transformations, vault, verification records, right-to-left
1.4.12026-10-01360 view across several systems
1.4.02026-10-01Remembered fields, undo, right-click, batch upgrades, templates
1.3.02026-09-30Labelled values, chained endpoints, result groups
1.2.02026-09-30Page watching in Auto mode
1.1.02026-09-30On/off switches
1.0.12026-09-30Captcha steps
1.0.02026-09-29First release

Version 4.2.0

Back to index ↑

Planned.

Less repeated work in batches, and quicker set-up. Includes everything in 4.1.0. Pro (the retry change applies to everyone).

Retry only the lookup that failed

  • When a value comes back partly answered (one lookup failed, the others answered), a batch retry now runs only the lookups that failed and keeps what the others returned. Before, the whole value, including sites and APIs that had answered, was run again. A site that failed is reopened from the top (login included); a site that answered keeps its tab. A value where nothing came back is still tried again whole.

Re-run the failed rows

  • On the scrape table page, Re-run the N failed in Batch… sends the inputs of the failed and partial rows on screen (each once) to the side panel's Batch tab and opens it where Chrome allows (otherwise they wait there). A panel that is already open takes them at once, switching to that profile.

Ready-made patterns

  • In a website's steps: + Sign in only if needed… adds the steps at the top (user name, password, sign-in button, an optional "wait for") that each run only when the sign-in form is on the page; the password must be a vault reference ({secret.name}), a typed password is refused. It can add the ↻ marker to the next step so a batch signs in once.
  • + Page through results… adds a "Repeat while" step that clicks Next and goes back to the step that reads a page, so the last page is read before the loop ends.
  • Both only assemble steps and conditions you can edit afterwards.

Row-number condition

  • "the batch row number …" is now a choice in every condition: is, is greater than, is less than, or is every Nth (the 1st, then every Nth after it), for example refresh the session or switch account on the first value and every 50th. A single run is row 1.

End-of-day reminder

  • In the profile's Batch, retries & table settings: "Remind me to download today's rows at HH:MM". A daily notification, on days when rows were kept, saying how many; clicking it opens the table page, where Download today saves them. Nothing is downloaded by itself.

Easier to find

  • A Scrape table button is always in the side panel's Batch tab (it used to appear only once a batch had rows), with a note saying whether the profile is keeping the table and where to turn it on; the note updates as you change the profile's settings. On the Free plan the note says it is a Pro feature.

Version 4.1.0

Back to index ↑

Planned.

Flow control for building one outcome from several websites and APIs, in single runs and in batches. Includes everything in 4.0.0. Pro.

Run only if…

  • Every website step, and every website or API lookup, can have a condition. A step can check the page (an element is on it, an element is not on it, an element's text equals/contains/matches…) or compare values built from tokens ({value}, a label, a constant, another lookup's {name.field}, a value set earlier): equals, does not equal, contains, empty, matches a pattern, greater/less than.
  • A step whose condition is not true is skipped (shown as skipped, with the reason) and the run carries on. A skipped lookup shows as skipped and the others still answer. Example: sign in only if the login box is on the page.
  • A lookup's condition may use another lookup's values, so it runs after that lookup.

Set a value

  • A new step that sets {name} from text with tokens, or from a formula when it starts with =. Later steps and conditions use it; other lookups read it as {website.name}.
  • "For the whole batch" keeps it across values while the website's tab stays open (needs a ↻ Repeat from here step) and forgets it when the tab is closed, so a "loggedin" value never outlives its login.

Repeat while

  • A new step that goes back to an earlier step while a condition holds, at most N times (default 20, 200 at most; 5,000 steps in total), optionally clicking something first (Next). A scrape that runs again labels its values Name [2], Name [3].

Continue in the tab of another website

  • A website can run in another website's tab, after it and after any lookup whose values it uses. One flow can now be: website A, an API using A's values, then A again with the API's answer, all in one tab. The shared tab is closed when the run ends.

Usability

  • The top button says what it will do: Run batch (N values) when values are waiting in the Batch tab and no single value is typed, otherwise Run.
  • Check now next to every page condition (and the Repeat-while condition): tries it on the web page you have open and says true or false, so nothing has to be run to debug "skip the login if…".
  • Pause and Continue in a batch, and a progress line: "5 of 20 · 4 ok, 1 failed · 1m 12s · about 3m left".
  • Open a failed row: a ▸ beside a failed (or partly failed) value shows the error, how many tries it took, and the steps of the last try.
  • Leftover tabs: if the side panel is closed mid-batch, the next batch closes the website tabs the earlier one kept open.
  • Scrape table page: click a heading to sort (numbers as numbers; click again for descending, a third time for newest first), a Columns menu to show, hide and reorder columns (remembered per profile, and followed by Copy, CSV and JSON; "Download today" always saves every column), click a cell to copy it.

Not built, on purpose

  • A jump to a label across lookups: lookups run as a chain of dependencies and a jump would break it and could loop forever. Conditions, repeat-while and continue-in-the-same-tab cover the same ground with limits.

Fixes

  • The Records tab always printed the text "null" under the list.
  • The new condition editor printed "null" the same way.
  • Conditions on steps now default to checking the page; conditions on lookups can only compare values.
  • The column headers of the batch table (side panel) and of the scrape table page were not over their columns: a style meant for the result cards also turned these tables into a grid. Fixed.
  • On the scrape table page, after deleting rows with a status filter on, the All button only cleared the dates, so the remaining rows stayed hidden until the page was reloaded. All now clears every filter (dates, source, status and search), and an empty view says how many rows are kept and to press All.
  • A stray "null" appeared under the batch table (and could appear in the Batch tab).
  • The Run button at the top now starts the batch when values are waiting in the Batch tab and no single value was typed (it used to do nothing useful, which made it easy to miss "Run batch"). Typing a value still runs that one value.

Version 4.0.0

Back to index ↑

In progress.

Major release for bulk scraping. Includes everything in 3.1.0.

Loop only some steps in a batch

  • A new ↻ "Repeat from here" button on each website step (Pro). In a batch, the address and the steps above it run once (a login, a navigation); the tab stays open and only that step and the ones below it run for each value, in the same tab. Values, variables and running numbers start fresh for every value.
  • A failed value closes the kept tab, so the retry starts again from the top, login included. Kept tabs are closed when the batch ends, stops or is cancelled.
  • A new Go to address step navigates the open tab (it can use {value} and other tokens, and respects the administrator's allowed sites), so each value can get its own page without signing in again.
  • Without the marker, batches behave exactly as before.

Wait, retry and stop when a site is blocked

  • Profile → Batch, retries & table: retries per value, wait between tries (fractions of a second allowed), "wait longer each time", pause between values, and "stop when a site fails this many values in a row".
  • Failures are counted per site, after its retries. The batch then stops and names the site ("“Registry” looks unresponsive or blocked: 5 values in a row failed there…"), and Resume carries on from there.
  • In a batch, an API that answers 429 or 503 with a Retry-After is waited for, once, before the value counts as failed.
  • A value where one lookup failed and another answered is kept as partial.
  • The side panel's Batch tab starts from the profile's settings (changeable for one batch) and shows the time each row was scraped.

The scrape table (Pro)

  • Turn on "Keep a table of every scrape of this profile" and every batch row (and, optionally, every single run) is kept with date, time, input, status, source and one column per value. Stored in the browser's IndexedDB, up to 50,000 rows per profile (oldest dropped first); sensitive values are never stored.
  • A full-page table (table.html, opened from the profile or the Batch tab): filter by date range, today, yesterday, last 7 days, source, status or text; Copy, CSV and JSON of exactly the rows shown; Download today for an end-of-day record; delete the rows shown.

Fixes

  • Deleting rows on the table page replaced its own "Deleted N rows" message with the row count.
  • A batch could not tell which lookup had failed (only a general error), so per-site counting needed the failing run's sections to be kept even when results are not saved to the side panel.

Version 3.1.0

Back to index ↑

Live on the Chrome Web Store.

Editing lookups is safer and faster. Includes everything in 3.0.1.

Clone a lookup

  • A ⧉ button on every website and API lookup makes a copy right after it, named “… copy” (then “… copy 2”), with new ids for its steps and values. The copy starts turned off, so the same system is not called twice until you have looked at it. Plan and policy limits apply.

Renaming no longer breaks the run

  • Changing a lookup's name used to leave every {oldname.field} pointing nowhere, so the run failed with "no turned-on endpoint is named …". Now Pluck counts the places that still use the old name (other lookups, checks, the Mapper) and offers to update them.
  • If you leave them, a ↻ button next to the name turns blue and updates them later, for example after importing a lookup and renaming it. The earlier name is remembered with the lookup, and exports and imports keep it.
  • The accordion title now follows the name as you type.
  • The same for a value's label: {lookup.label}, formulas that use it, the Mapper's "which value" choices and the "must be present" list.

API responses: add values quickly

  • Use on a value in the response now adds a new value labelled with its key, with the path filled in (before, it only copied the path). A copy button remains for the old behaviour. A path that is already a value is not added twice.
  • Join… on an object or list shows its children, lets you tick which to include, choose the delimiter (comma, space, semicolon, dash, new line or your own) and shows a live preview; it adds one value that joins them.
  • Paths can be put one after another with ++ ($.address.city ++ $.address.country), and a value can "Join everything it finds into one value" with a delimiter (also useful with [*]). Typed `

` means a new line.

  • The value's Test button understands both.

Version 3.0.1

Back to index ↑

Live on the Chrome Web Store.

Quality release for 3.0.0. Includes everything in 3.0.0.

Fixes

  • Sites named in the mapping are now part of the access Pluck asks for. Before, a second site could look "not open" and could not be filled until it was authorized by hand. Autofill now asks for any that are missing.
  • Free plan limit enforced: a profile holding several mapping pages (imported or hand-edited) filled all of them on Free. Free now fills the first page only, with no steps, and the status says so.
  • A value mapped to two sites no longer shows "field not on page" when the first site is closed and the second is open.
  • Stray null text next to the CSV and JSON buttons in the side panel when there was no run.
  • Undo data stays aligned when one tab fails during a fill.
  • The explanation under a page's Site box updates as you type.

Version 3.0.0

Back to index ↑

Live on the Chrome Web Store.

Major release: a page scanner and a Mapper, and Autofill across several sites.

Page scanner

  • Scan the current page for editable fields: text boxes, dropdowns (with their choices), checkboxes, radio groups, dates and rich text. Fields that already hold a value are included.
  • Skips read-only, disabled and hidden fields, passwords and card numbers.
  • Builds short, relative selectors (id, name, test id, aria, placeholder, label text, a short path from a stable container). Every candidate is checked to match exactly that field; the best one is used and the rest kept as alternatives (Pro).
  • Reads into iframes and shadow DOM. Understands tabs and sections: each tab becomes its own page with the step that opens it (Pro).
  • Watch for tab changes adds new fields as you open tabs; Walk the tabs clicks each tab header for you (after asking) and returns to your tab.

Mapper (Settings → Mapper)

  • Where each returned value goes is now set in one place, separate from the values themselves. Older per-value targets move into the Mapper automatically, unchanged.
  • Edit any selector by hand, Pick it on the page, or Test it; set which value goes in each field.
  • Order of events: a mapping is a list of pages, each one "do these steps, then fill these fields", so a flow like open tab A, fill, open tab B, fill, click Next is a list of pages. A page with steps and no fields (a Next or Save click) runs only on an explicit Autofill.
  • Move fields between pages, create a page from the move menu, and move a deleted page's fields instead of losing them.
  • Export and import the mapping on its own; it also travels with the profile.
  • Optional AI suggestions of which value belongs in which field (only if AI is already set up; sends field names and value labels only, never page content or returned values; cloud providers are asked every time).
  • Field kinds: dropdowns match by value or visible text; checkboxes by yes/true/1/on/checked; radio groups by value or label; dates in the page's format.
  • Free maps one page; Pro adds pages per tab or section, steps, watching and walking tabs, fallback selectors and AI suggestions.

Filling several sites at once

  • Each mapping page can have a Site (and optional Path), for example tms.example.com and /orders/*. One Autofill click fills each page in the open tab for its site, even when you are looking at another one.
  • Several tabs of one site: the tab you are on, otherwise the most recently used.
  • A site with no open tab is named in the status while the others are still filled.
  • Undo fill puts back the fields on every tab that was filled.
  • Automatic fill after a run uses the same logic.
  • Pages without a Site behave as before.

Other

  • Values on a tab that is not showing are marked "on another tab: Autofill opens it".

Version 2.2.0

Back to index ↑

Added

  • Export one or several website/API lookups from a profile to a file and import them into another profile, from a lookups file or from an exported profile. Constants, labelled values and vault secrets the lookup uses are carried or reported; clashing names are renamed; plan and policy limits apply.

Fixed

  • Retry on a section now runs the same logic as the Run button for that lookup: a captcha or "verify you are human" check is asked for in the side panel and the site's tab comes forward.

Version 2.1.0

Back to index ↑

Added

  • Settings → Alerts: scheduled-check changes can be posted to Slack, Microsoft Teams or any JSON webhook. The address is kept in the vault, sensitive values are never sent, and policy can switch it off.
  • Verification records can be printed or saved as a PDF report showing the systems consulted, what each returned, the hashes and the state of the whole chain.

Version 2.0.1

Back to index ↑

Added

  • Settings → General → Support → Copy debug report (no values or secrets).
  • AI provider keys follow the vault passphrase.

Fixed

  • A side-panel start-up error.

Version 2.0.0

Back to index ↑

Major release: multiple website lookups and enterprise features.

Lookups

  • A profile can have up to 5 website lookups (Free: 2) plus API endpoints. They can feed each other with {name.field}.
  • Independent lookups run together in background tabs; dependent ones wait; captcha prompts queue.
  • Old profiles migrate.
  • Secrets can be used in website steps.

Verification and monitoring

  • Selector health check.
  • Scheduled re-verification with change alerts.
  • Cross-system checks and a merged timeline.

Administration and setup

  • Enterprise policy through managed storage.
  • Remembered-fields editor.
  • AI "Describe it" to start a lookup from a description.
  • Oman Tax Authority template.
  • A compact "Add profile" menu.

Version 1.5.0

Back to index ↑

Transformations

  • Ordered steps and spreadsheet-style formulas on every value: text, regex, dates (including Hijri), numbers, currency, lookups and running numbers.
  • Live preview, constants, and reuse of earlier values as {name}.

Vault

  • API keys and tokens as {secret.name}, optional AES-GCM passphrase lock, never exported.

Verification records

  • A hash-chained, tamper-evident record per run, with JSON and CSV export.

Interface

  • Right-to-left layout setting. Profile sections are collapsible, with an Expand/Collapse all button.

Version 1.4.1

Back to index ↑

360 view across several systems

  • Sections view with per-system status and Retry.
  • Results stream in as each endpoint answers.
  • Table results for lists, a full-page view, record links.
  • Short-lived answer cache and sensitive values.
  • Customer 360 template.

Version 1.4.0

Back to index ↑
  • Remembered drop fields per profile and site.
  • Undo last fill.
  • Right-click and Alt+Shift+P to pluck a selection.
  • Batch upgrades: CSV loading with retries, stop-after and resume, copy and export of results.
  • Smart element picker.
  • Starter profile templates.

Version 1.3.0

Back to index ↑
  • Label several values on the source page and use them as {label} in endpoints, URLs and typed text.
  • Several webhook endpoints per profile, run in parallel, with {endpoint.field} chaining (Pro).
  • Settings → Groups: shared and per-profile groups, label assignment; Tabs or Cards view with a side-panel switch.
  • Target state shown as a highlight on the field name instead of a coloured card.
  • Free: 2 labels, 2 endpoints, 2 groups. Pro: unlimited, chaining, custom colours, per-tab Autofill.
  • Auto mode re-arms when profiles change; status is kept per tab and frame.
  • Auto mode: profiles with captcha or pause steps ask first, and the side panel shows why a watched change did or did not run.

Version 1.2.0

Back to index ↑
  • Auto mode can watch single-page sites for changed values without a reload.
  • Optional results area to watch, poll fallback, and a live "what the watcher sees" line.

Version 1.1.0

Back to index ↑
  • On/off switch for steps, scraped values, webhook values, query params and auth pairs.
  • Summaries show how many items are on (for example 2/3 values on).

Version 1.0.1

Back to index ↑
  • "Ask for captcha" step: shows the captcha and an input box in the side panel and types the answer into the page.
  • "Pause for me" step for reCAPTCHA or Cloudflare checks a person must pass.
  • "Run in the background" option: opens the site in a tab that is not switched to.

Version 1.0.0

Back to index ↑

First release.

  • Chrome MV3 extension that looks up values on a website, a webhook or an AI model and returns them in a side panel for drag or autofill.
  • Lemon Squeezy licensing (Free and Pro, with a trial).
  • Batch, history and Auto mode.
A little less busywork, explained

Meet Pluck in action.

See how a saved lookup turns everyday repetition into a little everyday magic.

Configure once. Open your page. Let Pluck help.Make yourself comfortable. Full screen is a click away.