JMX crawls your website, audits it against more than a hundred rules, measures its visibility to AI answer engines, and secures domain migrations — as a desktop application whose data stays on your machine.
For Windows, macOS, and Linux. The .NET runtime is included — install and get going.
Features
From the first crawl to a client report ready to sign — every feature runs locally, explainably, and without a forced subscription.
A spider in the spirit of Screaming Frog: follows links to the depth you choose, respects robots.txt and crawl-delay, reads XML sitemaps, follows redirect chains to the end, and evaluates every page on more than eighty signals — title and description including pixel width, headings, canonicals, hreflang, structured data, images, internal linking, response times.
More than a hundred checks, each with a German and an English explanation, a recommended action, and a link to the authoritative documentation — from missing titles through redirect loops and hreflang reciprocity to near-duplicate detection via SimHash.
Can an AI answer engine fetch, chunk, verify, and attribute the page? With an AI crawler overview and the distinction that matters: training crawlers versus answer crawlers.
OpenAI, Claude, Gemini, Mistral, and Grok assess crawled pages — titles, search intent, citability. Costs are shown before you start, a limit applies, and token usage is reported.
Before and after a relaunch: gather the inventory from a crawl, the sitemap, Search Console, or a site: query, map old URLs to new ones, assess every redirect, and weight the findings by the old URL's actual clicks. A broken URL with no traffic ranks below a broken one with a thousand clicks a month.
Crawl both environments and compare them page by page. Moved content is found across nine levels of decreasing confidence — from byte-identical source to an evolved slug — and emitted as ready-to-use server configuration: .htaccess, nginx, or web.config, with anything uncertain commented out.
Shop, blog, careers: “412 pages without a description” becomes the question of which area is affected. Whatever matches no definition stays visible under “No segment”.
A sitemap.xml from the crawl: all internal, indexable pages with status 200, as an index file from 50,000 URLs. A lastmod is deliberately not invented.
Check one or two sitemaps — as a URL or a file, an index together with all its parts — for everything the files themselves can answer: protocol and limits, every URL, lastmod, changefreq, priority, hreflang. Two inventories side by side show which pages only one of them lists, which only the other, and what both list in different spellings — the freshly generated one against the live one, before Google notices the difference.
Grep for your website: in the source or in the visible text, literal or as a regex, and also inverted — which pages do not have this. Runs on the loaded crawl.
Any value via CSS selector, XPath, or regex — prices, article numbers, social tags. A wrong selector does not cost a second crawl.
A crawl diagram with force layout, a crawl tree by click depth, a directory tree by URL structure — colorable by status, indexability, depth, or issues.
JMX fills in the login form and crawls with the session; CSRF tokens travel along unchanged. If the login fails, the run stops instead of silently mirroring the login page.
The draft applies to the crawled website instead of the live file — including its effect on AI crawlers. You see what it locks out before it goes live.
Presets for Googlebot, Bingbot, Applebot, AI crawlers, and real browsers — for the question of whether search engines get different markup than visitors.
Rendered against delivered HTML: which texts and links only exist after JavaScript — and which page sets its noindex only in the browser.
Apache, nginx, and IIS logs against the crawl: what is the crawl budget spent on, what has no bot ever fetched? Only what reverse DNS confirms counts as a bot.
Hunspell-based, each page in its own language, sorted by error probability instead of frequency. A second dictionary decides what is a loanword and what is a mistake.
Thirteen rules for what can be proven from the HTML — form fields without labels, icon buttons without names, duplicate IDs. With an honest note on what a static check cannot see.
Via the PageSpeed Insights API, with the distinction that matters: field data from real visitors separated from lab data for debugging.
OAuth sign-in, Search Analytics data, URL inspection for the real indexing status, sitemap management.
Licensed access to Google results for site: queries and index checks. The price of every call is shown, a daily limit applies first. No scraping.
Metrics, visit history, channels, and visits from AI answer engines as a channel of their own — whatever has been fetched once is never fetched again.
Which sections, in which order, under which heading — as a template per client, with logo, accent color, and introductory note. Export to HTML, CSV, and Excel.
Every tile and every row is a jump-off point: “412 pages without a description” opens exactly that rule, the URL opens the page, the arrow leads back.
A search box above every result list, literal or as a regular expression, and one matching field is enough: whoever types “contact” means the page, whether the word is in the address or in the title.
jmx-cli is the same tool for task schedulers, cron, and CI: crawl, audit, report, sitemap, exports — and --fail-on Critical ends the run red before your client does.
Download
All packages include the .NET runtime — nothing needs to be preinstalled. Choose your system:
If macOS claims the image is damaged: it is intact. The message stems from the missing Apple signature. Once, in the terminal: xattr -d com.apple.quarantine ~/Downloads/JMX-0.38.0-arm64.dmg
The checksums are available as a SHA256SUMS file next to the packages. Notes on SmartScreen and Gatekeeper are in the installation guide.
License
The key is derived from your details and the year of issue and is valid for the calendar year — until December 31. Misplaced it? Simply request it again with the same details: it is the same key.
JMX is provided free of charge!
JMX wird kostenlos zur Verfügung gestellt. Die Nutzung erfolgt auf eigene Gefahr: Die Software wird ohne Gewährleistung jeder Art bereitgestellt („wie besehen“), und eine Haftung für Schäden aus ihrer Nutzung oder Nichtverwendbarkeit ist — soweit gesetzlich zulässig — ausgeschlossen. Bitte prüfen Sie Befunde und Empfehlungen vor weitreichenden Entscheidungen selbst.
JMX is provided free of charge. Use is at your own risk: the software is provided “as is”, without warranty of any kind, and — to the extent permitted by law — any liability for damage arising from its use or unavailability is excluded. Please verify findings and recommendations yourself before making far-reaching decisions.