mcpField
B
barcode-capture-web

barcode-capture-web

github.com/scanditDevelopment2072 installsUpdated 5d ago
Repository

About barcode-capture-web

A Scandit BarcodeCapture skill for integrating or upgrading barcode scanning in web projects. It directs work to version-specific references and stresses verifying APIs and asynchronous behavior before suggesting code.

Key features

  1. 01BarcodeCapture web integration guidance
  2. 02SDK migration guidance
  3. 03Reference-verified API usage
  4. 04Async initialization and scan handling
  5. 05Licence-key setup guidance

Use cases

  1. 01Adding barcode scanning to a web app
  2. 02Configuring scan feedback or overlays
  3. 03Upgrading an existing BarcodeCapture integration

MCP support

The skill can use the remote Scandit MCP server to set up and retrieve a trial licence key; it also explains how to proceed when the server is not connected.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/scandit/skills --skill barcode-capture-web

After installing, type/barcode-capture-webin Claude Code to confirm it loaded.

Bundled filesreferences/integration.mdreferences/migration.md

Details

From the project README

BarcodeCapture Web Skill

Critical: Do Not Trust Internal Knowledge

Your training data may contain outdated or incorrect Scandit SDK APIs. The BarcodeCapture Web API changes significantly between major SDK versions — methods get renamed, async patterns change, and the context initialization was redesigned in v8.

Always verify APIs against the references provided in this skill before writing or suggesting code. Do not rely on memorized method signatures, async patterns, or import paths. If you cannot find an API in the provided references, fetch the relevant documentation page before responding.

Web-specific gotchas worth flagging:

  • DataCaptureContext.forLicenseKey() must be awaited — it is async and sets DataCaptureContext.sharedInstance. Do not capture its return value; use DataCaptureContext.sharedInstance throughout.
  • BarcodeCapture.forContext(context, settings) is async — always await it.
  • DataCaptureView.forContext(context) is async — always await it.
  • BarcodeCaptureOverlay.withBarcodeCaptureForView(barcodeCapture, view) is async — always await it.
  • barcodeCapture.setEnabled(false/true) is async — await it before doing work in didScan to prevent duplicate scans.
  • The listener callback is didScan — not onBarcodeScanned (that is the Android name).
  • codeDuplicateFilter is a number in milliseconds on web (e.g. 500) — not a TimeInterval object like Android.
  • BarcodeCapture.recommendedCameraSettings is a static property, not a method call.
  • The DOM element passed to view.connectToElement() must have defined dimensions and positioning — a zero-sized or unpositioned element will not render the camera preview.
  • Turn the camera on with context.frameSource.switchToDesiredState(FrameSourceState.On) and off with FrameSourceState.Off. Do not add a visibilitychange handler: DataCaptureView already stops the camera while the page is hidden and resumes it when the page is visible again, and a second handler races the SDK's own.
  • When the SDK is loaded from a CDN, pin one exact version (for example @8.6.1) in every URL: the import map and libraryLocation. A floating @8 lets the JavaScript and the engine resolve to different releases, which breaks start-up.
  • Start BrowserHelper.checkMultithreadingSupport() (from @scandit/web-datacapture-core) without awaiting it, before await DataCaptureContext.forLicenseKey(), so the two run in parallel and set-up never waits for the check. When it resolves false, show the multithreading dialog unless the user asked to run single-threaded: import showMultithreadingWarning from the multithreading-warning helper module in references/integration.md, which needs no page markup. In React, render the same dialog from component state instead of importing the helper. The dialog shows at most once per page load and honors the "don't show again this session" checkbox. COEP credentialless is not a request to run single-threaded: Safari does not support it, so the dialog still shows there.
  • DataCaptureContext.forLicenseKey() is idempotent, so calling it again is safe. dispose() releases the context, its engine and the camera: call forLicenseKey() again before scanning after it.

Intent Routing

Based on the user's request, load the appropriate reference file before responding:

  • Integrating BarcodeCapture from scratch, configuring settings, customizing feedback or overlay, adding a viewfinder, handling scans, or doing async work after a scan (e.g. "add BarcodeCapture to my app", "set up barcode scanning", "how do I use BarcodeCapture in web", "filter duplicate scans", "suppress the beep", "add a viewfinder", "disable scanning while I look up the barcode") → read references/integration.md and follow the instructions there.
  • Migrating or upgrading an existing BarcodeCapture integration (e.g. "upgrade from v6 to v7", "migrate my BarcodeCapture", "bump the Scandit SDK to v8", "what changed between SDK versions") → read references/migration.md and follow the instructions there.

API Usage Policy

Only use APIs that are explicitly documented in the Scandit references below. Do not invent or guess method signatures, parameters, or property names. If unsure whether an API exists or how it is called — or if a compile error occurs — fetch the relevant reference page before responding. Do not tell the user to check the docs themselves. After answering, always include the relevant link so the user can explore further.

Never construct or guess documentation URLs. When you need a specific class or property's API page:

  1. First check whether the page you already fetched contains a direct hyperlink to it — topic pages link directly to relevant API symbols. Always request links alongside content in your fetch prompt.
  2. If no direct link was found, fetch the API index (see Full API reference in the table below), extract the actual link from it, and follow that.

URL structures can vary (e.g. api/ui/ subdirectory) and guessing will lead to 404s.

Licence key

Scanning needs a Scandit licence key. For this skill the licence product is sdk and the licence platform is webassembly.

Work through these in order — never block the user on MCP:

  1. The Scandit MCP server is connected — call ensure_scanner_setup with product sdk and platforms webassembly, app_identifier (the app's bundle or application id; for web, its production domain, never localhost) and env_var SCANDIT_LICENSE_KEY. Do not run the command it returns. Then call get_license_key with the key's id, from that result or from list_license_keys; if list_license_keys shows the key is not a trial, ask the user before fetching it. Show at most a masked preview in chat; never print a full key.

  2. It is not connected — tell the user once that they can connect the remote Scandit MCP server, https://ssl.scandit.com/mcp, in their MCP client, then continue from step 1. Do not edit MCP client configuration yourself.

    Authentication is browser-based, so it is not supported headless or in CI. The server provisions trial keys only; it never creates, revokes, or modifies production licences.

  3. The user declines, or has no MCP client — they generate a key themselves at https://ssl.scandit.com (no account yet: https://ssl.scandit.com/dashboard/sign-up?p=test) and paste it in.

Then put the key into the code: write it into the licence-key call in place of the placeholder — the key get_license_key returned, or the one the user pastes or gives you — still showing only a masked preview in chat. Do not search the project, its files or the environment for a key.

References

Direct users to the right resource based on their question:

TopicResource
Get StartedGet Started · Sample
Advanced topics (viewfinders, location selection, feedback, duplicate filtering, composite codes)Advanced Configurations
Migration between major SDK versions6 → 7 · 7 → 8
Full API referenceBarcodeCapture API

Comments

Markdown supported