WP REST API Explorer — Tips & Tricks

Tricks for the WP REST API Explorer: reading findings right, the request limit and IP blocks, proxy pitfalls, login tests and pairing it with other tools.

Back to the overview: WP REST API Explorer · Open the tool: www.jpkc.com/tools/wp-api/

The manual explains every tab; the examples show the workflows. This page is about reading between the lines: how to weigh a result, where checks can backfire on you, and which combinations pay off. The interface is in English, so the real names appear here exactly as you'll see them.

Reading findings right

  • There's no overall grade — on purpose. A number like "87 out of 100" would offset a public user list against a Stripe key anyone can read. Instead, the Explorer counts per level. Read Action needed first, then Worth reviewing; Good to know is mostly WordPress default behaviour, which you can consciously leave as is or close.
  • Not checked is not OK. A check that couldn't run — limit reached, the site asked for a pause, bot protection, login required — appears with its reason under Not checked. Before you tell a client "all good", look at how long that list is.
  • confirmed or likely? confirmed means the tool saw it in a response, and the evidence shows it. likely follows from an indirect signal such as index data or field names. A likely is a reason to go and look yourself — not a certainty.
  • A sample doesn't prove absence. The run reads only a sample of each list. If the headline says "in the content we looked at", it means exactly that. For a solid statement about sensitive data, load the affected lists completely in Content with Load all items; the pattern scan picks up newly loaded data by itself.
  • "Nothing needs action right now." is strict. The Quick Check shows this sentence only after a complete run with everything loaded. With a sample or a partial run, the wording is deliberately more cautious.

Request limit and pace

  • 500 requests are enough for many small and medium sites. Large sites with many content types need more. If the run ends with "Results are incomplete: the limit of … site requests was reached.", raise the Request limit in Run settings and click Run remaining checks — the run picks up where it left off instead of starting over.
  • The limit covers the whole check, including your clicks afterwards. Single clicks may go past it, full pulls stop at the limit. If you plan several large pulls, budget generously up front.
  • Slower is sometimes faster. If a site responds sluggishly or asks for a pause via Retry-After, lower Parallel requests to 1 and Rate to 0.5 or 1. The tool never sends more than 60 requests per minute anyway. For large sites, the guide in the tool recommends starting the proxy with --timeout 60.
  • Every button shows its cost. "(1 request)", "(≈ 4 requests)", "(12 requests)" — the Explorer tells you before a click what it will send to the site. Afterwards, the request log shows every single request.

Don't get your own IP blocked

  • P1 and P3 are the touchy probes. The request for a user number that doesn't exist and the ?author= redirect look like a user-name scan to security plugins. That's why the tool sends both at the very end, warns about them explicitly in the Security tab, and drops them if the site asked for a pause earlier.
  • Switch probes off when it's delicate. If you check someone else's production site with strict protection, turn off Include probes in Run settings. The affected checks then appear under Not checked; you can send individual probes later, on purpose, with Send this probe (1 request).
  • If the run pauses with "The site has started refusing requests", wait a few minutes before you press Resume. If the diagnosis says "Your IP may have been blocked by the site's firewall (user enumeration protection).", the only cure is waiting — or allowing your IP in your own site's firewall.
  • A 403 isn't necessarily WordPress. The Reference tab says it: a 403 can also be a firewall's HTML page, and 502/504 come from the proxy, not the site. The diagnosis tells these apart; the request log shows the status of every request.

Proxy pitfalls

  • The token changes on every start. After a restart, the tool reports "The token does not match" or, with 1.4.0, that the proxy could not prove it is genuine. Copy the new token and you're done. If that gets old, start with --token-file — but be aware that a stolen token then stays valid across restarts until you delete the file.
  • The first connection always fails. When the browser asks about the local network, the request that triggered the question is lost, even after "Allow". Nothing to worry about — the tool checks again shortly afterwards, or you press Check again.
  • A double-click starts on a random port. On Windows, create the start-localproxy.cmd from the guide, or copy the Address line from the proxy window into the Proxy address field.
  • "… was started for a different site" means the proxy runs with a different --origin, for example one for another website or a local copy of the tools. Restart it with the command from the guide; Copy start command builds the right one.
  • Keep --allow-local as narrow as possible. *.ddev.site allows every DDEV project on your machine. According to the proxy documentation, "read only" for local targets relies solely on the tool's code — any script on an allowed origin that holds the token could write to them too. So only run the proxy with --allow-local while you're testing.
  • DDEV and DNS: don't use --dns system for DDEV names; routers with rebind protection won't answer *.ddev.site at all otherwise. Intranet names, on the other hand, need --dns system. And for HTTPS, mkcert -install on the proxy machine is a must.

Making sense of plugins, theme and version

  • The real plugin count is almost always higher. A plugin only shows up if it leaves a trace — a file on the home page, a REST namespace, a content type. Backend plugins without a front end stay invisible. WordPress gives the complete list only to administrators, so only with a login.
  • "One of several plugins" is honest, not indecisive. Some signals are shared by several plugins. Rather than guess, the tool counts the group as one plugin.
  • The version from API features is a lower bound. "at least x.y" comes from features the API offers; a generator tag, if present, shows the exact number, and the site can remove or edit it. If signals contradict each other, the tool says so. It deliberately makes no claim about whether a version is outdated or insecure.
  • Don't run unknown plugin routes blindly. In Structure, the tool warns before Run anyway for good reason: even a GET route can trigger something on the site. The tool sends only GET and HEAD (and with a login the proxy refuses anything else too), but neither can prevent side effects of reading routes.

Login tests with care

  • Use a test account with few rights. The login checks only judge roles with few rights: the abilities check roles that cannot write posts (such as Subscriber), the extra-data check Subscribers and Contributors. An administrator mainly shows the complete plugin and theme list.
  • Delete the application password afterwards. The tool forgets credentials when you reload; in WordPress, however, the application password stays valid until you remove it.
  • 401 despite the right password? After Apply, the tool lists the likely reasons under "Likely causes": a wrong user name or application password, a host that strips the Authorization header before WordPress sees it (the usual rewrite rule in .htaccess helps), a separate HTTP Basic Auth password protection in front of WordPress — or, on an http site, that WordPress turns application passwords off without HTTPS.
  • If the site redirects, the login is gone. Credentials apply only to the origin you entered them for. If http:// redirects to https:// or to www, enter them again for the new address.

Handling files with care

  • Exports contain personal data. The tool masks passwords and keys as far as it recognises them; names and e-mail addresses from the site stay in JSON, CSV and snapshot files. Treat the files accordingly.
  • "Show passwords and secret keys in full" is for your eyes only. The tool spells it out: "Only for your own use. Do not pass this file on." The same goes for snapshots that include the site's original answers.
  • For clients, use the customer report. It names no routes, no names and no evidence. The technical report masks contact details and keys, but it does show author short names.
  • The download protects you from yourself. If your proxy token or your login ends up in a file, the tool refuses to save it. What you can share is the link from Copy link — it only fills in the address for the recipient, who needs a proxy of their own.

Pairing it with other JPKCom tools

  • WP Code Generator — generate the snippets behind Open the matching snippet there as a must-use plugin. Apply them, repeat Check site and use Compare against the old snapshot to show what's been fixed. Prefer a ready-made plugin? JPKCom Disable XML-RPC switches XML-RPC off globally, and JPKCom Harden & Trim bundles hardening measures for WordPress.
  • SEO & GEO Analyzer — rates HTTP security headers and TLS, which the Explorer explicitly doesn't judge, plus the SEO side of the home page.
  • DNS, SSL, Redirect & URL — if the diagnosis reports a certificate problem or "The site redirects in a circle.", trace the chain and redirects there hop by hop.
  • JSON Editor — for the JSON export or a single API response from Structure: browse it as a tree, filter it with Transform or validate it against a JSON Schema.
  • CSV Editor — review and touch up the export's CSV tables.
  • Markdown Editor — open the reports as .md, add your own notes for the client and save the result as a PDF.
  • localproxy — Guide & Tips — background on downloading, starting and the endpoints of the proxy; what the Explorer needs from it is in the manual.