Node.js 26 Compatibility

• Support TLS connections to IP addresses on Node.js 26 without sending an invalid SNI hostname.

Sitemap list, and more ways to start a dig - October 11, 2026

Added

• Open Current Tab in Digger and Open Clipboard in Digger, two commands that dig the active browser tab (macOS, with the Raycast Browser Extension) or the URL on the clipboard. Both are off by default.
• View Sitemap is now a searchable list. Search covers every entry's full text, not only the rows shown, and a sitemap index loads its child sitemaps so search reaches their pages too, up to a size limit; the list says what it didn't search. Open a child sitemap to browse it on its own, or dig any page in it with ⌘↵. Sitemaps are cached for 48 hours; ⌘R reloads.

Changed

• Digger's URL is now optional. Left empty, it digs the URL in the selected text, then on the clipboard (both on by default), then in the current browser tab (off by default). These settings existed before but never applied, because the URL was required.
• Text that isn't a URL, such as a single word, is now shown back as "isn't a URL" instead of being dug as a website and failing to connect.
• When a dig fails, the search bar shows the URL that failed. Correct it and press ↵ to dig the new one.

Fixed

• A sitemap lastmod date such as 2026-08-07 no longer shows as the day before in time zones west of UTC.

SVG extraction, and faster access to Well-Known files and color tokens - October 1, 2026

Added

• View All SVGs in Resources & Assets — the SVGs in a page's HTML: inline, sprite symbols rebuilt as standalone files, <img>, <object>, favicons, and in-page CSS url(). Copy as SVG, Copy as PNG on macOS, or Export as SVG.
• Preview SVGs on a white, black or gray backdrop (⇧⌘B) so logos drawn in black or white stay readable, and Quick Look any SVG with ⌘Y.

Changed

• Image thumbnails and the site's favicon now download through Digger's network check before they are shown, so a page cannot make Raycast request an address on your local network.
• The Images, Fonts, Stylesheets and Scripts blocks in Resources & Assets each show their own icon — image, text, swatch, code — green when the dig found some and red when it found none.
• Pressing Enter on the Well-Known and Theme sections now opens their file and token lists, instead of reopening the site you just dug in a browser, which did nothing specific to either section. Open in Browser is still in the action panel.

Well-Known files, Theme colors, and resource exports - September 11, 2026

Added

• Well-Known section — probes 82 paths under /.well-known/ from the IANA registry and reports which ones a site actually publishes, with the contents one keystroke away. A directory cannot be listed, so each path is asked for individually and judged by its opening bytes rather than its status code: a single-page app answers 200 to everything.
• Theme section — the colors a site declares: theme-color per media query, color-scheme, the data-theme family, and the CSS custom properties from its stylesheets. oklch(), lab(), color(display-p3 …) and var() chains are all resolved so each one shows a real swatch, viewable as a list or a color grid.
• Copy and Download actions on robots.txt, sitemap.xml, and every well-known file — as text, Markdown, or CSV, with the sitemap's URLs and robots.txt's directives expanded into rows.
• Copy actions in Metadata for the page title, description, Open Graph description, and Open Graph image URL.
• Code Formatting preference — pretty-prints minified JSON and XML when viewing a file. Copy and Download always give you the original bytes.

Fixed

• Pages served in the wrong language. Digger sent no Accept-Language, so sites that negotiate picked for it — one returned Arabic for an English page, and every field read from that page was the wrong translation.
• sitemap.xml reported as malformed when the site simply has none. A 200 serving an HTML error page now reads as absent, and a 406 as unavailable, instead of "could not parse".
• Long lists ending in a clipped ......ore instead of "…and 9 more", in Resources, HTTP Headers, Discoverability, and Data Feeds.
• Translucent colors rendering as fully opaque swatches.
• A sitemap containing an out-of-range character reference such as � preventing the sitemap view and its export actions from opening.

Security

• Stylesheets are fetched only from the site being analyzed or a public host, with every redirect re-checked. A page could previously name a stylesheet on localhost, a private network address, or a cloud metadata endpoint and have Digger request it. They are also now read with a hard byte cap rather than buffered in full.

Report failed DNS, certificate, Wayback and host-metadata lookups - September 3, 2026

Added

• The Overview row now shows the site's own favicon as soon as it decodes, falling back to a globe rather than sitting on a chain icon while a third-party favicon lookup runs.
• Host metadata (/.well-known/host-meta) now appears under Data Feeds & API. It was being fetched and cached on every dig but rendered nowhere, so the result was invisible even when it succeeded.

Changed

• Updated to Raycast API 2.x.
• A lookup that fails now says so in its own section — DNS, SSL certificate, Wayback and host metadata each read "Couldn't check" — instead of a single truncated banner row at the top of the list. A toast announces it once. Because the statuses travel with the result, a cached dig still explains itself.

Fixed

• A failed DNS, SSL certificate, Wayback Machine or host-metadata lookup rendered an empty section with nothing saying why. Those failures are now listed in the "Some data couldn't be loaded" banner, with the underlying cause.
• A dig that came back empty is no longer cached. A site answering an unhappy status with no body — a bot-block or an edge error — parsed into a result with no title, resources or metadata, and caching it pinned that empty view for the full 48-hour TTL, so re-running the command changed nothing.
• A non-2xx status is now marked in Overview. Only 200 carried an indicator, so an error status read like a normal result next to an empty page.
• "Refresh" now actually refreshes. It read straight through the cache, so on a stale or broken result — exactly when a user reaches for it — it re-rendered the same result and looked like the action did nothing. It now re-digs the site, bypassing the cache. Other digs still use the cache, and a failed refresh does not discard the cached copy — re-running the command still finds it.
• The Markdown report claimed to include TTFB and a request count, but neither was ever measured, so those lines could never appear. The dead branches and the unpopulated fields behind them are gone.
• A DNS lookup that failed outright now says so on every record row. It added a "Couldn't check" line at the top of the section while the six rows beneath it still read "No IPv4 addresses found", "No mail servers found" and so on — absences from queries that never returned.
• Each DNS record type now reports its own outcome. A host can answer for one type and fail on another — an A record found while the MX query times out — and the failed row said "No mail servers found", a claim about the host from a query that never returned. Those rows now read "Couldn't check".
• A DNS section that couldn't be checked no longer reads as a host with no records. This covers every way the lookup can fail, not a handful of named ones — a malformed resolver response or an out-of-memory error now reads "Couldn't check" like a timeout does. If the first record type came back empty — normal, plenty of hosts publish no AAAA or MX — and the resolver then failed on everything after it, only that first harmless result was kept and the section reported "no records found" for a lookup that never completed.
• The Wayback Machine section no longer invents a snapshot count. When the precise-count request failed — and archive.org times out often — it fell back to a formula meant for huge archives, reporting a site with 8 snapshots as 5,000, labeled an estimate. It now says "Couldn't check".
• A Wayback lookup is bounded and cancelable. Its four sequential requests each retried independently with no shared limit, so an unresponsive archive.org could hold the section for over a minute; and a dig you replaced kept fetching for a result nobody would see. The whole lookup now shares one budget and stops as soon as its dig is superseded.
• robots.txt, llms.txt and sitemap.xml no longer report "Not found" when the check itself failed. A 5xx, a timeout or a refused connection now reads "Couldn't check" in that row, and rows read "Checking…" until the request comes back. Only an answer from the server — a 404 or a 410, or for robots.txt and llms.txt a page that is really an error page — says the file is absent.

Failed digs no longer hang, and shortcuts work on Windows - August 11, 2026

Fixed

• A dig that failed outright — an unreachable host, a mistyped domain — left the spinner running forever and never showed an error. It now reports the failure immediately.
• Errors identify the actual problem. A DNS failure now reads "Connection Failed" with connection-specific advice and the underlying cause, instead of a generic "Fetch Error".
• Keyboard shortcuts now work on Windows. Every custom shortcut previously used a ⌘-based binding that does not exist there, leaving those actions unreachable.
• In the HTTP Headers list, "Copy Header Name" and "Copy Header Value" were both bound to ⌘ ⇧ C, so one of them could never be triggered. "Copy Header Name" now uses ⌘ ⌥ C.
• Digging a new URL before the previous one finished could let the abandoned dig overwrite the new results, blank its panels, or fill its progress bars.
• DNS & Certificates no longer shows the previous site's certificate when the current site's TLS lookup fails.
• "Open in Wayback Machine" and "Save to Wayback Machine" now open the correct app when running Raycast beta.

Added

• Copy Error on the failure toast and Copy Error Details on the error screen, both copying the same report: the URL, the underlying cause, and the suggestions.
• Copy Error Details on the "Some data couldn't be loaded" banner, for the individual component failures.

Changed

• The failure screen is now a centered empty state rather than a list row with a detail pane beside it.
• "Save to Wayback Machine" moved from ⌘ ⇧ S to ⌘ ⇧ Y, and "Copy Canonical URL" from ⌘ ⌥ C to ⌘ ⌥ U, so they no longer shadow Raycast's standard shortcuts.
• URLs written to the debug log are stripped of their query strings.

Add Content Signals and Payment Required (x402) detection - February 19, 2026

Added: Content Signals detection

• Digger now parses Content-Signal directives from robots.txt and displays them in the Discoverability section

Added: Payment Required (x402) detection

• Digger now detects x402 payment-required signals from HTTP responses and surfaces them in two places:
• Discoverability section: primary indicator showing which signals were found (HTTP 402 status code, PAYMENT-REQUIRED header, PAYMENT-RESPONSE header)
• HTTP Headers section: supporting detail listing the raw values of x402 protocol headers
• Payment Required signals are included in the Markdown report export (⌘ ⇧ M)

Added: Favicon display

• Digger now displays the favicon as the Overview icon when available (thx @jlokos for #1)

Improved: URL processing

• Enhanced URL extraction and normalization across all input sources (argument, clipboard, selected text, browser extension):
• Improved extractUrl() to handle trailing punctuation (e.g., https://example.com. → https://example.com)
• Added validation to extracted URLs to ensure only valid URLs are accepted
• Fixed priority order: bare domains like example.com are now preferred over embedded URLs in mixed input
• Browser extension tab URLs now benefit from the same extraction logic as other sources
• normalizeUrl() now properly lowercases scheme and hostname via URL parsing

Changed: Screenshots and dependencies

• Updated screenshots
• Updated dependencies

Initial Version - January 27, 2026