Every WorkBOS product update, security fix and improvement, versioned and dated as it ships: the running record of how the business operating system evolves.
Why that is the wrong test right now. The alarm began by returning early whenever nothing was waiting and nothing had been given up on, under a comment reading "mail is moving; nothing to say". Nothing waiting also means nobody has tried to send anything. An empty post box tells you nothing about whether the post box works. WorkBOS has not started marketing, so quiet stretches are the normal condition rather than the exception. In September the queue only grew because digests are generated on a schedule; on a slower day the same dead account would have gone unreported for longer than forty-five hours, and the first person to find
What changed. There is now a third reason to raise the alarm: nothing can send at all. It is checked against the sending account itself rather than inferred from the queue, so a dead account is reported within the hour whether or not anyone has written to us. The message says plainly that nothing is waiting and nothing can go out, names when sending last worked, and still carries the cause in words rather than an error code.
What it deliberately does not do. It does not ring the moment a sender stops being ready, because a sender that fails once waits a short while and comes back by itself, and an alarm for that is noise, and noise is how an alarm gets ignored before the day it matters. It requires that no sender has succeeded for three hours, the same tolerance the piled-up-mail branch already uses. It also stays quiet on a platform that has never sent anything at all, because that is somebody halfway through setting up a sender, and nagging them about the thing they are in the middle of doing is not an alarm.
WorkBOS now notices a broken email connection by checking the connection, not by waiting for messages to pile up behind it. Before, a sending account could fail quietly for as long as nobody happened to be sending anything.
Nothing in WorkBOS changes. This records an outage this evening and what caused it.
A new guide answers the question most agency software marketing avoids: does the product keep your books, or does it send your numbers to Xero or QuickBooks? Seven products, sorted by that one question, including where another tool is the better choice.
The screen could not tell you what it knew. Setting up Gmail sending needs two values from Google: a client ID and a client secret. The secret was handled properly, with the box saying that one was stored. The client ID was never shown at all, so an empty box looked exactly the same whether a value was saved or missing, and a client ID is not a secret: it travels openly in the query string of every Google permission screen, so there was never a reason to hide it. Worse, the screen decided whether to say "a secret is set" by checking both values together, so with the ID missing it also reported the secret as gone. Of the two, the secret
And the deletion left no trace. Every change to a sender writes an audit row naming which fields changed. The list named the API key, the SMTP password and the Google client secret, and did not name the client ID, so the one field whose disappearance took all outbound mail down for 45 hours was the one field whose disappearance was not recorded. That is why the cause had to be reconstructed from function bodies two days later instead of read off a screen.
What changed. The two values are asked about separately. The client ID box shows what is stored and says plainly when nothing is. The secret's own message depends only on the secret. If Connect refuses for a missing client ID while the secret is present, it says so, says where to copy the ID from, and tells you the secret does not need re-entering. And a change to the client ID is now named in the audit trail, by name only, never by value.
Setting up Gmail sending needs two values from Google, and the screen used to show neither clearly: a missing one looked the same as a saved one, and losing the first made the screen claim the second was gone too. Both are now reported honestly, the one that is safe to display is displayed, and a change to it is written into the audit trail.
Two fixes to signing in and to email. If you have more than one Google account, the sign-in screen now offers "Use a different Google account" instead of always using the one your browser last used. And the screen for reconnecting our email sending is repaired. It used to clear the setting it was meant to renew.
WorkBOS can now be listed in the official registry of MCP servers, which is where AI assistants and developers look to find one. Nothing in the product changes; this is the paperwork that proves the listing belongs to us.
What changes. The recording now hangs off the one piece of code that every way in already passes through, so sign-up, sign-in, password resets, failures and blocked attempts are all captured, and a new route added next year cannot forget to do it. Each record keeps the country, region, city, postcode, timezone and connection address, read straight from our hosting provider at no cost and with no outside service to be blocked by.
Mailboxes are now matched properly. `t.ariq@gmail.com`, `tar.iq@gmail.com` and `tariq+trial@gmail.com` are one mailbox and were three customers. For a product about to give a plan away free, that was the whole problem in one sentence: one person, unlimited free workspaces, no tools required. They now collapse to a single identity, and a new panel lists any mailbox or network holding more than one workspace.
Nothing refuses anybody. This release only watches. You chose that deliberately, and it is right: with very little sign-up traffic there is nothing to hold back, and setting the thresholds from real traffic beats guessing at them and turning away a real customer.
Two things found while building it. The sign-up capture would have failed silently for a second time, for a different reason, and the check that was supposed to prove it was writing did not exist; it does now, and it can tell "nothing is happening" apart from "something is broken". Separately, the check that keeps our published erasure figures honest could not see three tables whose names contain dates, so /compliance and /security would have told customers we erase 328 tables when we erase 331. Both fixed.
A third fault, found by checking rather than assuming. Before shipping, the record of what the live database actually looks like was refreshed, which is a step this project requires precisely because a stale copy quietly agrees with itself. It showed that five of the new pieces of the database had been left readable by anybody on the internet: not the data, which is locked to us alone by a separate rule that was in place and working, but the ability to ask the question. The instructions to lock them were written down correctly and had simply not taken effect when the change was first applied. They are applied now, and the live database has been re
The erasure register covers the new table. `identity_events` is on it as personal data with no safety copy taken, because a restore point of where people connected from is not a thing worth keeping. With the three date-named tables now visible to the checker that had been missing them, the plan reads **331 wiped / 42 exempt**, and **248 of the 331 are backed up, 83 are erased and deliberately not**. Those are the figures /compliance and /security publish; until this release they published 328.
WorkBOS now records which country, city and connection each sign-up and sign-in comes from, so you can see where your customers are. Nothing is blocked or refused by this. It is the groundwork for offering free and discounted plans without them being abused.
What changed. Before creating anything, sign-up now asks whether the product can actually send. If it cannot, nothing is created and the person is told the truth: we cannot send email at the moment, this is at our end, try again a little later. Their address stays free for a real sign-up afterwards. Password reset is deliberately left alone, because reset links come from the sign-in service's own mailer and still work.
If WorkBOS cannot send email, signing up now says so instead of creating an account and telling you to check an inbox nothing will arrive in. Your email address stays free, so you can sign up properly once sending is working again.
Also in this release, several places where our own pages sold the free plan as something it is not. The pricing page said "the whole product on the free plan" and "five seats, the whole product" directly above a table marking the ledger, payroll and voice as Pro. The homepage description sold "projects, CRM, HR, payroll and real double-entry accounting" under the words "free for 5 users, no card", and five comparison pages did the same. The free plan runs projects, CRM and invoicing. The general ledger and payroll are Pro at $12 per user per month, and every one of those sentences now says so.
Connecting Claude, ChatGPT or Cursor to your workspace did not work: assistants were sent to an address that could not answer them. That is fixed, and connecting now behaves the way the /mcp page describes. Separately, several pages described the free plan as including the accounting and payroll. It does not, it never did, and those pages now match the pricing table: free covers projects, CRM and invoicing, and the general ledger and payroll are on Pro.
Housekeeping inside our own build process. Nothing in WorkBOS changes.
AI assistants often answer questions about WorkBOS with details about WorkOS, WorkBoss or WorkBoard instead, because those names are close and those companies have been written about far more. Our site now publishes a clear list of the accounts that genuinely belong to WorkBOS, including the LinkedIn page and the YouTube channel, so an assistant can tell which is which. Nothing inside WorkBOS changes.
What changed. There is now an alert inside WorkBOS itself (the same kind that already warns when a backup goes stale) that appears when mail stops moving. It rings on anything waiting more than three hours, and on anything given up on in the last day, which matters because unsent mail eventually expires and an alarm watching only the queue length would fall silent exactly as the failure became total. It says how many, how long, and why in plain words: a withdrawn Google permission is named as one, and an error it does not recognise is passed on as written rather than flattened into something nobody can act
WorkBOS could not send email for two days this week and had no way to tell anyone, because the way it tells you things is email. It now raises an alert inside WorkBOS when mail stops going out, saying how many messages are waiting and why, so this cannot go unnoticed again.
Three new comparison pages are live: WorkBOS against Workamajig, Scoro and Productive. These are the products AI assistants recommend when somebody asks for agency software with real accounting built in, and until today we had no page answering any of them. Each page says where the other product is the better choice as well as where we are, and every price on them was read off that company's own pricing page this week. Nothing inside WorkBOS changes.
Every browser that opens WorkBOS is handed a key that is signed by that secret. It is in the live JavaScript, as it is supposed to be, and it is what lets a visitor reach the sign-in screen at all. Change the secret and every one of those keys becomes invalid at once: the product stops answering for everyone until a fresh version is published. That is not a rotation, it is an outage we scheduled ourselves.
The way out is one value. The key the app hands out can be a newer kind that is not signed by anything, so the secret no longer holds anything hostage. Nine files read that value and not one of them looks inside it, so it is a single change rather than a rewrite.
Proved rather than assumed. A one-value change that every browser depends on deserves evidence, and evidence somebody can re-run instead of a paragraph in a message. There is now a check that calls both keys against the same five things the product actually does - a public form recording a visit, a stranger being refused a customer's data, a call reaching a background function, the sign-in screen loading, a private file store refusing a stranger - and compares the answers. Run today: **identical on all five.** It is written so that once the switch is made it reports "the migration is complete" and stops, rather than fail
Ready, and deliberately not done tonight. It changes how every browser identifies itself, so it goes live with you awake rather than at two in the morning with nobody watching. It is four steps, the last of which is finally rotating the key, and going back is the same one value and about three minutes. The steps are written down in full.
Nothing you can see. Preparation for replacing a security key, including a check that proves the replacement behaves identically before anything changes.
Every release carries a version number, and the What's new page shows you which one you are running. The tool that hands those numbers out could give the same number to two pieces of work, and in one case could have let a release publish under a number belonging to somebody else's work - so the version you saw might not have matched the build you had. Both are fixed. Nothing else in WorkBOS changes.
Your workspace decides who may delete what - some records only by the person who created them, or by an owner or admin. Those rules were being applied everywhere except the Delete button itself, so a team member could remove a client, contract, form, job post, lead, offer or proposal created by someone else. That is fixed, and the message now tells you which rule stopped you and who can help. Deleting your own tasks is unchanged. Deleting a project is now owners and admins only, which is what your rules already said everywhere else.
Every credential record would have been lost. `snrpmo.secrets` refused to load - all 10 rows, which is every credential the platform holds: Stripe, both phone carriers, Google, GitHub, LinkedIn, X, TikTok, Supabase, Vercel. Not their values, because backups deliberately never contain those, but the **record that they exist**: names, accounts, which provider, when each is due for rotation. In a real disaster that record is the checklist you rebuild the company from, and it would have come back empty with no warning.
This was the same defect found in July on two different columns. It was fixed for those two and the rule was never written down, so it grew back on a third. The rule is now enforced against the whole database rather than against a list of names, and the check is mutation-tested, deliberately broken to confirm it fails when it should, because the first version of it silently matched nothing and passed.
The drill also corrected me mid-fix. My first attempt made one of the two columns nullable and reasoned the other was safe. The re-run moved the same error one column along, in about ten minutes. That is the entire argument for running the thing rather than reasoning about it.
The backup restores. 138,269 of 138,459 rows came back. The 190 that did not are eight hand-made safety copies somebody created directly in production before three plan changes in September; no migration ever created them, so a rebuilt database has nowhere to put them. They are now reported by name rather than counted as a failure, because a check that can never pass is a check everyone learns to ignore, and they are a concrete argument for the decision already waiting on you: **those copies cannot be restored, so they are not protecting anything.**
Nothing you can see. We restored a real backup into an empty database and checked it, and fixed a fault that would have lost the record of every connected account during a recovery.
The list of bullets under each release on the changelog was showing the heading and dropping the sentence under it, so a bullet said what changed but never what it meant. Nine releases also carried a bullet called "Under the hood" with nothing after it, which is a heading from our own engineering notes and should never have been shown to you. Both are fixed, so from now on every bullet carries its explanation, and anything with nothing to say is left out.
This one is the owner's decision, so it has not been changed unilaterally.
No change to what you see. Two internal release checks that had been failing were investigated and settled: free internal calling is confirmed working and correctly included on Pro and above.
What actually happened.
What was wrong.
So it can be seen next time.
What it found on its first run.
Platform administrators: the Platform menu now shows a number when something needs attention, the Activity page says whether an action was taken by the platform or by a customer on their own workspace, and the Security tab lists who has signed in and what admins did in the last week.
The roles editor lets you decide who may add, change and remove email addresses, mailboxes and your workspace's email settings. Those switches were being saved but not enforced, so an admin you had restricted could still do all of it. They are enforced now, and both Email screens ask first before offering a button. Nobody in any workspace currently has one of those switches turned off, so nothing changes for you today - they simply work from now on. Your own mailbox is not affected: removing or disconnecting the account you connected yourself never needed an admin permission and still does not.
What changed.
WorkBOS was down for about twenty-five minutes on 17 September while the daily backup was being built. Nothing was lost. The backup can no longer run twice at the same time, which is what tipped the server over, and the checks that run before a release now wait for a backup to finish rather than running beside it.
The practical lab.
Renewal that keeps your serial.
The Advanced level.
For the platform desk.
The Academy now has a practical lab: press one button, get a free practice workspace of your own, and do ten real set-up tasks that WorkBOS checks for you, task by task. Passing it is the Hands-on verified mark and is now part of the Certified Specialist certificate. Certificates renew in the 90 days before they expire, keeping the same serial, and a new Advanced Specialist level is awarded automatically to certified specialists with a passed lab and three or more verified client ratings averaging 4.0 or better.
Request to hire.
Badges that are earned, not typed.
Trust signals on every card.
Keeping it clean.
You can now hire a certified specialist from inside WorkBOS: send a short brief from their page, get an answer within 72 hours, and invite them into your workspace with one click. Every specialist carries a badge they earned (Certified, Verified Pro, Top Rated), their availability, and how reliably they answer.
This release changes nothing you can see, and it is listed because of what was found rather than what was shipped. Deleting a record goes through Trash, and Trash was checking only some of the rules your workspace sets about who is allowed to delete what. The fix for that was written, tested against real data, and then taken back out within two minutes because it would also have stopped your team deleting their own tasks - which is a change to how WorkBOS works rather than a repair, and not one to make without asking. Everything behaves exactly as it did before.
A fresh paper every time. A module check draws 12 questions at random from a bank of 30, the exam draws 40 from 120, and the answer choices are shuffled, so two attempts never look the same. The fourteen banks are live now - 390 module questions and 120 exam questions, each tagged with the topic it tests and the lesson it comes from.
Time that runs on our clock. A check is 15 minutes, the exam 60. Your answers save the moment you give them; when the time is up, what you have answered is handed in, and a browser that died mid-attempt loses nothing that was saved. Twelve hours between attempts as before, and at most three attempts in any seven days for a check, two in thirty for the exam.
The exam, from the Certificates page. It opens only once every other requirement is met, and it begins with an honour pledge. Afterwards you see a score by area, never the answers. A module check you did not pass names the topics to revisit and links to the lesson to rewatch.
Recorded, not judged. The Academy records how an attempt was sat - how often the tab lost focus, how long it stayed hidden, whether anything was pasted, and the time taken against everyone else's. Anything that stands out is flagged for a person to look at, never punished by a machine. A flagged exam pass puts the certificate application on hold until someone has read it.
Module checks and the final exam now draw a fresh set of questions for every attempt, save your answers as you go, run on a clock, and open the exam only when everything else is done. The Academy records how an attempt was sat and a person looks at anything that stands out before it counts towards a certificate.
Bank statement import is the one thing deliberately left alone.
A permission in the roles editor that governed nothing.
If your workspace is on the Free plan you can now do the part of invoicing that was quietly broken: mark an invoice paid, take a payment that settles several invoices at once, raise and apply a credit note, and turn an accepted quote into an invoice. All of that is included in Free and has been listed on the plan all along; the system was checking for a different, paid feature before letting you through. Six workspaces were affected. Importing a bank statement genuinely does need the Financial plan, and that screen now says so rather than opening and then refusing.
You are no longer asked to approve your own request.
A portfolio rename that quietly undid itself.
If your access to a workspace is read-only, the app no longer shows you buttons it was always going to refuse: the phone dialer, scheduling a meeting, starting a project from a template and sending a text message are simply not offered any more, rather than turned down after you press them. You are also no longer asked to approve your own request, which somebody else has always had to decide. And renaming a portfolio you manage but whose company you do not used to appear to work and then quietly put the old name back; that button is now only shown to the people who can really use it.
"Other provider" now adds your own address, not the company's.
Eight pages under Administration used to open on a plan that did not include them.
Four more buttons that were never going to work have stopped pretending.
The email screen reads properly on a phone.
Your email settings now tell you the rules before you act on them: if your workspace only allows its own domains, the page names them above the Connect Google button, and if it has not confirmed one yet, it says so instead of letting you hand Google permission for an address that could never send. "Other provider" adds your own address rather than a shared one, and editing your own address works again. Twelve buttons across the app that your plan or your role was always going to refuse now either explain who can use them or are simply not there, and the email screen reads properly on a phone.
Overview.
Database.
Jobs & queues.
Security.
Errors & incidents.
Releases.
Platform administrators: the health page is now seven tabs - what needs attention, database, jobs and queues, backups, security, errors and incidents, releases - each opening with its verdict and the actions that go with it. Off-site backup copying, which had been silently refused since 5 September, works again and now says plainly that a destination still has to be chosen.
Demo learners, one button each.
Release A watched through, live.
The Academy console can now show you how the first pass works before anyone has learned anything: add four demo learners from Platform > Academy > Settings, look at the Integrity ledger and the module pages, and remove them again in one press. The learning path itself was watched through live on a real lesson and one small glitch at the moment of completion is fixed.
The trim was being overridden by the stylesheet.
"Long" is judged by the answer, not the question.
Every cell opens the record.
Proved on screen.
The daily backup had been failing since this morning.
On a form's submissions page, long answers now really do show as two lines with the full text on hover, so a pasted essay no longer stretches the row. Clicking anywhere on a row opens the submission. The same trim fix reaches chat search results and Notes.
Saving your sending email works again. A database change on 9 September removed a key three other parts still relied on. Every save raised an error for a week.
A check that looks at every page before a release. Real pages, desktop and phone, light and dark. It found six things no existing test could see.
The dashboard stopped hiding your work. Needs attention said 16 items and showed four. Two cards could come up completely blank.
You can remove a mailbox, not just disconnect it. Disconnect keeps the setup so reconnecting is one click. Remove deletes the connection and everything synced from it.
The Mail tab explains why Gmail is greyed out there. Sending as your address works today. Reading your inbox needs far wider permission and is not built.
Saving your email settings works again. It had been failing since 9 September and we are sorry. You can now remove a connected mailbox, not just disconnect it. The dashboard no longer hides items behind a fixed height or shows you a blank card. A few labels that were placeholder text or the wrong size have been corrected. And there is now a check that looks at every page before a release, which is what found most of this.
Analytics that answer a question.
Click a form submission to open its own page: every answer in full, the lead it created beside it, and reply by email, call, WhatsApp, follow-up task, spam and correct-the-answers one press away. The Analytics tab now says whether people finish the form, whether you reply fast enough, which source turns into customers, and when submissions arrive.
The first pass goes in order.
No skipping ahead, no faking.
"Still with me?".
The rules, and the ledger.
The Academy now teaches the way a good course does: lessons in order, no skipping ahead on the first pass, "still with me?" prompts and quick questions, a lesson complete when it has really been watched, and fair-play rules everyone agrees to. The platform desk sees what was refused and can reset a lesson.
Disappearing messages.
Send later.
A switched-off feature is now visible where you look for it.
Delete a message for yourself or for everyone, with a minute to undo and, for the deleter, whether anyone had read it. Disappearing messages on groups and direct conversations. Send later: pick a time and WorkBOS says it for you.
A red dot that counts everything, everywhere.
A popup when a message arrives, set up the way you want it.
One reaction per person on a message, a red count for unread messages on the sidebar, the tabs, every chat, group and discussion, and a popup for new messages that you set up your way: which ones, with or without the text, which corner, how long, sound.
In the Calls tab, tap a call and its details open right under it instead of in a side panel.
A call history that reads like a phone's.
Delete a conversation from your list (the other person keeps theirs, and it comes back if they write again), delete a group you manage for everyone, or select several and delete them together. The Calls tab now reads like a phone's call log: grouped by day, with what happened and for how long, full details on a tap, Call back, and delete on each entry or several at once.
Fixes the crash on opening a conversation that v1.318.0 did not fully close, and the broken profile picture in Team Chat.
"Available" now means WorkBOS is open, not that you are looking at the Team tab.
Your own status, at the top of Team Chat.
Every person in Team Chat now carries a status dot: green available, grey away, red do not disturb, amber in a meeting, blue on a call. Hover to read it; in a direct conversation the words are under the name with the time a meeting or a do-not-disturb ends. Set your own Do not disturb from the top of Team Chat for an hour, four hours or until tomorrow, and it clears itself. Available now means WorkBOS is open on any page.
Everyone in the room answers in place, once, and can change their mind until you close it.
Who, what and when, as a dashboard and a sheet.
A form you already have can be asked in a room too.
Ask a group, a project room or a colleague for structured answers without leaving the chat: a drop-down with an "other" box, a number, remarks and more, built behind the paperclip next to Poll. Everyone in the room answers once, can change their answer until you close it, and the answers open as a summary dashboard and a sheet of who said what and when, with CSV. A published form you already have can be asked in a room too. Who may answer is decided by who is in the conversation.
The crash report now leaves the phone.
Fixes v1.317.0: opening a conversation with messages in it crashed the page. Every conversation opens again, on phone and desktop. Crash reports from the error page now reach us.
The tabs are Chats, Projects Discussion, Team and Calls.
Every message has a menu.
The whole emoji set, voice messages, polls, location, a colleague's card, and a screenshot pasted straight in.
Pin, mute and colour your conversations; filter the list.
Pictures open in WorkBOS's own photo viewer, never in a new tab.
Share into a conversation from your phone.
Team Chat opens on the newest message and loads older ones as you scroll. Tabs are Chats, Projects Discussion, Team and Calls; colleague calls are separate from the phone system, with your own history and owner-controlled clearing. Every message has a menu: reply, forward, copy, edit, pin, save, mark unread, seen by, delete. The whole emoji set with search, voice messages, polls, location, colleague cards, paste a screenshot, drop a file, bold and italic. Pin up to five conversations, mute, colour and filter them. Pictures open in WorkBOS's own photo viewer with save, edit, rename and forward, never in a new tab. Share into a conversation from your phone's share sheet. Owners get a Rules screen for the whole workspace's chat.
Deleting a message in a channel works again.
Groundwork for the next release, already in the database.
A fix for channels and groups that would not load or send since this morning's release, and a repair to deleting messages in channels.
Attach anything, from inside WorkBOS or outside it.
Mentions that read like names, and reach agents and Chief.
Reply, react, and an emoji picker that closes.
One message box for everything.
Six people who were never real.
A question only a member may ask.
Team Chat has groups: make one from people or from a team, set who can post, who can edit and who can add, mute or pin it, and see every change written into the room. Attach files, links, Drive files, projects, tasks, invoices and more to any message. @ mentions show names instead of codes, work from the keyboard, and can reach your AI agents and Chief. Reply with a quote, react in any room, and the emoji picker finally closes when you click away.
Calling a colleague no longer opens the customer phone dialler.
Call history belongs to the conversation, not the phone system.
Better sound on internal calls.
Read and delivered marks.
Emoji and reactions.
The conversation list finally tells you something.
Team Chat is now one place: your chats, your project channels, and everyone you work with, in three tabs. Calling a colleague opens a proper call panel instead of the customer phone dialler, and closes itself when the call ends. Calls with a colleague now appear inside your conversation with them rather than in the phone system's log. Messages show Sent, Delivered and Read marks, you can react with emoji, and the conversation list shows the last message with a readable time and an unread count. Internal calls now use echo cancellation and noise suppression at both ends. On a phone, the chats list works properly for the first time.
The Message button went to the wrong place.
The page looked thin, and it was.
Three views of the same people, and you pick.
Filters for the way you actually think about your team.
The Message button on Colleagues now opens your conversation with that person instead of the channel list. The page itself has been rebuilt: everyone appears as a card with their job title, department and the local time where they are, and there are now three views - Cards, a full sortable List you can export, and a Board you can arrange by department, company, team, status or access. Filter by department, company or team, and search by name or job title.
It knows when someone is in a meeting, which the big chat tools cannot.
The workspace owner decides who may contact whom.
Two holes closed while building it.
Private messages are private, unless the owner says otherwise - and then everyone is told.
Calling is built and deliberately switched off.
Deleting a workspace reaches the new records too.
There is a new **Colleagues** page in the sidebar. It shows everyone in your workspace, whether they are free right now or on a call or in a meeting until a given time, and gives you a message button beside each name. There is a do-not-disturb switch at the top. If you own the workspace you can group people and set rules for who may contact whom. Reading private messages is off unless you turn it on, and if you do, everyone in the workspace is told and every read is logged. Calling a colleague is built and will be switched on once we have tested a real call end to end.
Adding a domain now really does look it up. Type the name and WorkBOS fills in who it is registered with, when it was registered, when it expires and its nameservers. This was announced in the previous update but did not work on the live site; it does now.
Start free and every release lands in your workspace the moment it ships.