JSOPanel Get started

Release notes.

What changed in JSOPanel, one version at a time. Newest first.

v0.1.287 21 October 2026
  • Fixed: a site behind the automatic protection from AI scraper bots, with its own custom per-path rewrite rule set up for it by request, could have the wrong page answered for a visitor - the rule's own rewritten path reached your real site instead of what was actually asked for. Existing sites with this shape pick up the fix automatically on your next update.
v0.1.286 21 October 2026
  • Fixed: the "who's using the most right now" overview from the previous release never actually showed any accounts on a server using the newer resource-limit layout, which is most servers - it was only ever checking the older one. It now checks both, in the same order the rest of the panel already does.
v0.1.285 21 October 2026
  • Added: Resource Usage now shows, for anyone looking after more than one account, which accounts are using the most processor, memory and disk activity right now, each with today's highest reading and a short line showing the last few minutes. You are emailed when an account actually runs into one of its own limits - not merely getting busy - with no repeat email for the same ongoing incident within half an hour.
v0.1.284 21 October 2026
  • Added: the server now watches for a database query that has been running for an unusually long time and emails you about it, naming the account, how long it has been running and what it is doing - switched on by default, with an optional setting to have the server end a query like that automatically once you are comfortable turning that on.
  • Added: a hosting plan can now set a maximum number of database connections and queries per hour for every account on it, with a recommended starting point already filled in, so one account under heavy load cannot slow the shared database down for everyone else.
v0.1.281 20 October 2026
  • Added: the automatic bot filter can now let ChatGPT, Perplexity and Mistral's own traffic through when they fetch a page live for one of their own users, rather than refusing it the way every AI-related visitor is refused by default. Checked against each company's own published list of addresses before it is allowed, so a visitor merely claiming to be one of them is still refused - and the separate traffic those companies use to train their own models is still refused either way. A real server-wide switch (Security Center's own Visitor check tab) decides what any site with no choice of its own does, and each site can still set its own choice - always let it through, always refuse it, or follow the server's own setting - from that site's settings page, administrators only, once the bot filter is already switched on for that site.
v0.1.280 19 October 2026
  • Added: an AI assistant connected to the panel can now list, add and remove a customer account's SSH keys - adding one can also turn SSH access on for that account if it is not already, the same rule already used when a reseller is allowed to do that themselves.
  • Added: an AI assistant can now read a domain's error log, access log, or its own firewall (ModSecurity) block history.
  • Added: an AI assistant can now change a domain's PHP version, or its PHP settings such as memory_limit.
  • Added: an AI assistant can now read a file from inside a customer account's own home directory - read-only.
  • None of the above can ever be used on the server's own root account - that stays something only you can do, signed in directly. Each capability needs its own permission ticked on the assistant's token first, the same as every other assistant action, and is off by default.
v0.1.279 19 October 2026
  • Fixed: a site using the automatic protection from AI scraper bots could send the same security response header twice in one reply. Harmless to most visitors, but it could confuse a strict security scanner or another server reading the response. Existing sites pick up the fix automatically on your next update.
  • Added: an AI assistant connected to the panel can now sign a customer straight into their own account - the same "Login to control panel" link a billing system can already offer.
  • Added: an AI assistant can now check whether a customer's own address is blocked on the server.
  • Added: an AI assistant can now release a customer's own blocked address - the same "Unblock my IP" action already available to a billing integration.
  • Each of the three above needs its own permission ticked on the assistant's token first, the same as every other assistant action.
v0.1.277 18 October 2026
  • Fixed: the five extra security response headers (on by default, every domain) were never actually reaching a site that was being answered from a second/standby server, a suspended account's own notice page, or the Mailman mailing-list archive - only an ordinary, directly-served site got them. All three now carry the same headers everything else does, and existing servers pick this up automatically on the next update.
  • Added: a site can now be exempted from part or all of the five extra security headers on its own, without switching them off for every other site on the server - useful when a site genuinely needs to be embedded in a frame on another page, or embed one of its own. Set it from that site's own settings page, Security response headers for this domain (administrators only).
v0.1.276 18 October 2026
  • Fixed: the previous release's fix for domain-name lookups failing inside the web terminal or an SSH session did not actually correct an account that was already showing the problem - the repair believed it had replaced the broken file, but nothing had changed for an account already in that state. It now genuinely replaces it. If the previous release did not clear this for you, this one does, with no action needed on your next update.
v0.1.275 18 October 2026
  • Fixed: a hosting account using the web terminal, or signed in over SSH, could not look up domain names from inside their own session - tools like `curl`, `git` or `getent` would fail with "could not resolve host", even though the rest of the account worked normally and the server itself had no DNS problem at all. Every account with shell access was affected. Existing accounts are corrected automatically on your next update, with no action needed.
v0.1.271 16 October 2026
  • Fixed: on a site using the automatic protection from AI scraper bots, a setting meant to remember that a visitor had already proven they were a real person - so that proof would survive the protection restarting - had never actually worked, because of an internal naming mistake. In practice this meant every visitor who had already passed, on every site using it on a given server, lost that pass at the same moment, whenever the protection restarted for any reason. It is corrected, and the fix applies itself automatically on your next update.
v0.1.270 16 October 2026
  • Fixed: a visitor who correctly answered the "Just checking your browser" check could be shown it again roughly every hour, over and over, for as long as the server kept running - even though they had already passed. The server now catches up with everyone who has passed within a few minutes, instead of only when something unrelated happened to prompt it to.
v0.1.269 16 October 2026
  • Fixed: transferring an account from another server, then telling that server to stop publishing its domains, could report success even when a domain had not actually been released - with nothing afterward to say so. A domain left behind this way could, in rare cases, be mistaken for one still genuinely hosted here, which could interfere with DNS management on a server using DNS Cluster. A release that does not fully finish is now reported as not done, with a standing notice on the Transfer page naming exactly which account, which server, and which domains still need releasing - and a domain is now checked against this server's own records before it is ever treated as one of its own.
v0.1.268 16 October 2026
  • Changed: adding, removing or leaving a DNS Cluster server now says plainly when that action couldn't complete because the server's own DNS settings failed to reload - instead of a message written for a different situation that gave no hint it was connected to what you'd just pressed.
v0.1.267 13 October 2026
  • Fixed: a "Send to a program" address could silently stop delivering in three different ways - "keep a copy" ticked with no real mailbox behind it, that mailbox deleted later, or the account renamed - and nothing said so, so it was only ever found from a bounced message. All three are now caught: the mismatched setting is refused when you try to save it, turned off automatically if the mailbox it depended on is removed, and warned about if a rename can't carry it across.
Page 3 of 5

Want to see it running for yourself?

Get in touch and we'll walk you through it.