UScraper May 2026 Product Update: Smoother Canvas, Workspaces, Multi-Text Loops, and Loop Continue

Smoother drag on the canvas

The canvas was the first thing we touched. Moving blocks that already sit on the canvas now feels much steadier: the block captures the pointer, stores the starting pointer and block coordinates, compensates for the current zoom level with dragScale, and updates its position through translate3d(...). ScraperCanvas batches those position updates through requestAnimationFrame, so fast pointer movement no longer floods React with one state update per browser event.

A few things changed under the hood:

• Pointer events and pointer capture keep the block in control until the drag ends, even if the cursor moves quickly.
• Zoom-aware coordinates divide pointer delta by canvas scale, so dragging feels right at 50%, 100%, or 150% zoom.
• requestAnimationFrame batching lets the canvas apply the latest pending position once per frame instead of re-rendering for every move event.

If you've ever dragged a block across a crowded graph and felt the motion lag behind the cursor, that's the bug class we went after.

Workspaces for parallel projects

Heavy users have been telling us the project list grew faster than the sidebar could keep up. Workspaces are the answer: each Workspace is a named container with its own project list, browser sessions, and run history. Switching workspaces keeps your canvas zoom, the open browser, and recent run state intact for that context — and leaves the others untouched.

Typical setups we're already seeing internally:

• Research workspace for review-site scrapers (Trustpilot, G2, Capterra).
• Lead Gen workspace for people-finder and directory pipelines.
• E-commerce workspace for listing and price-monitor projects.

The new sidebar workspace list keeps recent projects scoped per workspace, adds a New Workspace flow with color selection, and lets projects move between Standalone Projects and any Workspace through the row menu.

Multi-input Type Text

Navigate has supported a Multiple URLs field for a while — paste a list and the block runs once per URL. The same pattern now exists on Type Text via a new Multiple Texts parameter. Drop in a list of search terms, filter values, or any string and the downstream graph runs once per value, reusing the same browser session.

A few details worth knowing:

• One value per line in the Multiple Texts textarea. Empty lines are skipped.
• clearFirst still works — the field is cleared before each value so you don't end up concatenating keywords.
• The run panel shows the current iteration (2/4) and the current value in the run-step label, so you can tell at a glance which input is being processed.
• It's purely opt-in: a Type Text block with a single value behaves exactly like before.

This is ideal for "run the same flow for each keyword" workloads — SERP sweeps, marketplace category dumps, multi-region searches — without duplicating block branches in the graph.

Loop Continue, an explicit boundary

Multi-input loops on Navigate could quietly restart by closing and reopening the browser, but that trick doesn't generalise to other blocks: a session opened mid-flow, cookies set during login, or scroll state can't be recreated for free. So we added a new Loop Continue block to make the loop boundary explicit.

Here's the mental model:

• Loop Continue advances the innermost active multi-input loop and jumps execution back to the block that started it (the Navigate or Type Text with multiple values).
• When the loop is exhausted, Loop Continue falls through along its outgoing edge so the run can continue toward export, an End block, or a parent loop.
• End is unchanged: it still terminates the entire run and closes the browser. It does not advance loops.

The runtime keeps active loops on a stack, so nested multi-input loops work the way you'd expect: an inner Loop Continue only advances the inner loop, and the outer loop only steps forward when its own Loop Continue runs.

A worked example

Picture a small SERP harvester:

1. Navigate to https://example.com/search.
2. Type Text with laptop, keyboard, monitor, webcam in the Multiple Texts field, targeting the search input.
3. Click the search button, Sleep briefly for hydration, then Structured Export the result rows.
4. Go Back to clear the page state.
5. Loop Continue so the next keyword starts cleanly from the search results page.
6. After the last keyword, Loop Continue follows its outgoing edge into End.

The same browser session stays open for all four keywords, the export file appends across iterations, and the run panel labels each step with the keyword and the iteration index, so debugging stays grounded.

Frequently asked questions

This update is rolling out automatically — restart UScraper to pick it up. We'll keep iterating on Workspaces, internal loops, and canvas ergonomics through the rest of the quarter, so let us know what you'd like next.