Everything here already shipped.
Newest: Your agents have their own email addressesEach agent on your team now has an email address of its own. The new Mail page shows every inbox as conversations, so you can read what came in and what your team said back. When an agent writes an email, it waits inside that conversation for your approval before it is sent.
Open MailYour team now writes a long video script like an article: built around one real story, in the words you would say on camera, with a note on how a viewer is likely to take it. Shorts and Reels follow one real case and are written to be read aloud. Your team learns how you talk from your own video transcripts, in the language you speak on camera, even a mix like Hinglish.
On workspaces running the latest version, agents now work a website the way a person does: open a page, click, type, and look at what changed before the next step. That lets them get through sign-ups and pages behind a login, and you can still watch the session from chat and take over. When you ask an agent to sign up for a site, it creates the account itself instead of asking you to save a login.
Open ChatSquadRoom now has a blog at squadroom.app/blog, about working with an AI team: how to brief it, why each task has one owner, and what an AI employee actually does. Each post opens with a short summary, and you can copy it as text or open it in ChatGPT or Claude to ask about it.
When an article covers your own product, your team now runs a free public test of your pages and shows real screenshots of the tool, instead of reusing old numbers. Articles link to your own site where it fits. Comparison pages read like a buyer's guide, and your team tries both products before comparing them.
Open DocsThe model picker in Settings shows the model that is really chosen, even when it is not one of the newest. Asking for changes on a document that waits for your approval now sends the work back right away. In chat, agents can again look at your tasks and mail.
Open SettingsSquadRoom now has help docs at squadroom.app/docs: short pages on getting started, your workspace, your team, settings, and what to do when something goes wrong. Problem pages quote the exact message you might see, what causes it and how to fix it. You can search the docs, and copy any page to paste into an AI assistant.
When your credits run out, research pauses, and a task that needs it waits in the queue with a note on its card. It starts again by itself after you buy more credits or your plan renews. A playbook you stopped can now be set up again, and it keeps its notes and tables.
Open TasksStarting a workspace over now lets you pick what stays. Your connected apps and AI providers are kept unless you say otherwise, and you can keep your team's memory too. When you make a backup, you choose whether it includes your saved logins.
Open SettingsYour team now keeps what it learns in one shared memory and checks it before each task. It picks up rules from what you say in chat, comments and answers, and each one quotes your own words. On the Memory page you can see everything it keeps, add your own notes, and remove anything.
Open MemoryEach task now shows its plan: a short list of steps, who takes each one, and how far it has got. Your team writes the plan when work starts, ticks steps off as it goes, and for some tasks asks you to look it over first.
Open TasksAgents on your team now ask each other for help while they work, and each task shows who did which part. When one agent finishes, the next step goes to the agent who owns it, and your approval can start it: approve an article, and it can be added to your website on a set day. An agent can also split a big task across up to 12 helpers, up from 2.
Open TasksBefore your AI team starts, you now fill in a short form about your company: its name, website, what it does and who it is for. Your team researches from there, including any apps you have connected, and the setup chat asks only what it could not find. Each new team is also asked for a small first piece of work within about an hour, so you see it working on day one.
Questions waiting for your answer, reports, and the outreach messages your team drafts for you are now written in plain, everyday words. Task threads are shorter too: agents post fewer comments, each one brief, and long ones fold away until you open them.
Open TasksYour agents can run on your Claude subscription instead of an API key, right from the first step of signing up. You can disconnect it or swap it for a key at any time.
New workspaces that use OpenAI now start on GPT-6.1 Sol. If your account cannot use it yet, your agents fall back to GPT-5.6 Sol, and a model you already picked stays as it is.
Open SettingsWhen a teammate sends a draft back for changes, the writer now picks it up again right away, and a few rounds of feedback no longer pause the task as if it had failed. If a piece of work does stop, the larger task it belongs to now says so on its card.
Open TasksSetup now asks about more than who buys from you: why people pick you over competitors, where you want to be in three months and in a year, and what good work looks like to you. Your new team starts with a real picture of the business and your taste. You can skip any question or ask to speed things up.
Each blog topic your team suggests now says where it came from: something people search for, recent news, a competitor's page, or your own story. Your team checks what your blog already covers so it does not pitch the same topic twice. Social posts are now drafted every day for that day's conversation, instead of a week at a time.
When an agent asks a question with a few likely answers, the options now show up as buttons. Tap one, pick several, or type your own reply as usual.
Open ChatThe Docs page now opens on what your team delivered to you. Research sits on its own tab, one note per subject, marked with how long ago it was checked. Your team updates an existing note instead of researching the same subject from scratch.
Open DocsDocuments no longer end in a long list of links. Sources and key details now sit in a side panel next to the text. For articles and landing pages, the panel also holds the page title, description and link slug, each with a copy button, ready to paste into your website.
Open DocsArticles your team writes can now include screenshots of a product's page and a related video. In Docs, a YouTube link on its own line plays in place, and an uploaded PDF opens in a preview you can download.
Open DocsPause one task, one agent, or the whole team, and pick up again when you are ready. You can also set a monthly credit limit on the Usage page: you get a warning at 80%, and at 100% the team pauses until you raise the limit or the new month starts.
Open SettingsA new Passwords & Logins tab in Settings keeps the website logins your agents need, so they can sign in to those sites while they browse for you.
Open SettingsType / in an agent's chat to pick a command, such as setting a goal, making a plan, scheduling work or checking status. Skills you have installed show up in the same list.
Open ChatWhen an agent is using a web browser, chat shows the session live. Press Take Control to drive it yourself, for example to sign in to a site, then hand it back when you are done.
Open ChatAgent replies in chat now come as a few short messages, like a texting app, instead of one long block.
Open ChatTweet Tester previews a post in an X feed and estimates its likes, replies and views. Is Agentic scores how ready a website is for AI agents. AI Slop Detector shows which sentences in your writing sound like AI, and AI Squad Planner turns two lines about your business into a plan for an AI team. All four are free and need no account.
OpenAI, Anthropic, Google and Azure do not show what is left on an API key. The Usage page now shows what your workspace spent through each of those keys in the last 30 days, priced from the providers' published rates.
Open SettingsA message you write while an agent is still replying now waits above the chat box and goes out when the reply ends. You can also send it right away, edit it, or move it to the front.
Open ChatEach task now has its own web research budget. When it runs low, agents switch to cheaper ways of reading the web and finish with what they have. Repeating a lookup you already paid for no longer costs again.
A document can now be a table with columns and rows, like a simple spreadsheet. Your team keeps lists of posts, articles and prospects there, and you can edit cells, add rows and columns, and filter, search and sort. Every save is kept in the history, so you can go back to an earlier version.
Open DocsStarting a workspace over no longer throws away the AI accounts and keys you connected. They come back connected, with the order you set them in, and the reset screen lists what is kept and what is deleted.
Open SettingsOn the Memory page, Import into shared memory gives you a prompt to run in ChatGPT, Claude or another assistant you use. Paste its answer back and your team keeps the facts worth remembering.
Open MemorySkill Scanner checks an AI agent skill for risky instructions before you install it. The Skills Marketplace lets you browse public agent skills, on the website and inside the app. Paste shares any Markdown at a link that can expire. All free, no account needed.
When your team asks to publish a post, reply or email, you now see the exact text and where it will go. If a connected account can send it, one tap approves and sends it; otherwise copy it with one click. Some requests offer a few versions so you can pick the one you like.
Open DashboardEvery document now has a Request changes button. Say what should change and it goes to the agent who wrote it, or to a new task if that work is finished. You see where your request will go before you send it.
Open DocsEvery request waiting for you now has a Chat button. It opens a chat with the agent who sent it, with the work already attached, so you do not have to explain the context again.
Open ChatYour team can now work with about 30 more apps, including Salesforce, QuickBooks, Asana, Jira, Airtable, Zendesk, Mailchimp, Zoom, Calendly, Dropbox and Figma.
Open IntegrationsYou can flag any agent comment, chat message, report or deliverable to tell the SquadRoom team a result was wrong. One tap sends the item along with your note.
A browser session now costs 10 credits. You get a warning when your balance drops to 15% and again when it runs out, and the Usage page shows which agent spent what.
Open SettingsThe Calendar now shows each recurring job on the days it actually runs, including monthly and custom schedules. Past runs show whether they finished or failed.
Open CalendarWork your team suggests now has its own page that you go through one card at a time: start it or dismiss it. The dashboard's "Needs you" count now shows only the things actually waiting on your answer.
Open TasksPlaybooks are ready-made recurring jobs you can give your team. They now end in finished work, such as a draft or a reply, instead of a summary memo. A new daily Social Pulse drafts posts only when there is something worth saying.
Open PlaybooksWhen your team asks to set up a recurring job, the approval now says in plain words what it does, how often it runs, and which business number it should move. Schedules run on your workspace's clock.
Open CalendarThe task board now separates work that is running from work waiting on you or on another task. A task that is stuck names what it is waiting for.
Open BoardIcebreak writes opening lines for cold emails from what is publicly known about a prospect. It is free and needs no account.
Every running task card now shows a short line saying what the agent is doing at that moment, in plain words, such as "Pulling the 3 overdue invoices from the books". It updates live and disappears when the work stops.
Open TasksModel pickers now show a short list of each provider's current models, with a recommended one first, instead of every model the provider offers. The full list is one click away.
Open SettingsEach playbook run now ends in a clear next step: a decision for you with a recommended option, a few suggestions for follow-up work, or a short note that nothing needed doing.
Open PlaybooksWeek Plan proposes up to five tasks for the week for you to approve. Unblock Sweep finds work that stopped moving. Ship Log turns finished work into release notes. Customer Voice finds what customers keep asking for and suggests one change.
Open PlaybooksEach task now shows what happened and when, in order, so you can follow it from start to finish without digging through logs.
Open BoardGive any agent a new name whenever you like. Older tasks and chats still point to the same agent.
Open SquadA comment on a task in progress now reaches the agent working on it, so you can change direction mid-task instead of waiting for it to finish.
Open BoardThe launch plan now comes with 1,000 credits. If you run out mid-task, the work stops with a clear message and a link to top up.
Open SettingsYour team lead can now pass work to the right specialist, and specialists can suggest work they think is worth doing. Nothing they suggest starts until you say yes.
Open BoardAdd a new specialist to the team you already have, without rebuilding it.
Open SquadIf you run your own MCP server, a standard way to give AI agents tools, you can now add it from Integrations and your team can use its tools.
Open IntegrationsSee a whole month of scheduled work at once. One job can run on several schedules, and you can open any past run from the day it ran.
Open CalendarMentions and new requests for you now collect in one inbox, so you can see what is waiting on you in one place.
Open DashboardEach step an agent takes, including the web pages it reads, now shows up in chat as it happens. Open any step to see exactly what was done.
Open ChatPick from a library of templates for common recurring work instead of writing instructions from a blank page.
Open SkillsChat now has a separate channel for each agent, and conversations are kept between visits.
Open ChatYour team can send you a brief each morning: what happened and what is waiting on you.
Open ChatInstead of filling in a form, you tell your team lead what is taking up your time. It asks a few questions and proposes a team of AI agents for you to confirm.
Skill Evaluator scores how useful an AI agent skill is and flags risky instructions, with a result page you can share. Gauntlet Loop writes a prompt that keeps a coding agent reworking its output until it beats a real example. Both are free and need no account.
For each app you connect, choose whether your team can only read from it or can also take actions in it.
Open IntegrationsOne search box finds docs, tasks and calendar items by what they mean, not only the exact words, and jumps to any page. Open it from the top bar or press Ctrl+K (Cmd+K on a Mac).
Agents can sketch a flow or a layout on a whiteboard that opens in chat, and you can edit it.
Open ChatWhen your team thinks of more work to do, it now suggests it instead of starting on its own. Your team lead filters the list first, and you approve or decline what is left.
Open BoardSet a timezone for your workspace and scheduled work runs on your local clock.
Open CalendarA new Usage page shows how much AI your team's work used, so you can see where the spend goes.
Open SettingsEach run of a recurring job now appears on the task board like any other task, so you can follow its progress and results.
Open BoardBring in skills you already have by uploading a file or pasting a link.
Open SkillsShare any document with people outside your workspace through a read-only link, and turn the link off whenever you want.
Open DocsYour agents can run on another AI provider, you can connect more than one account, and switching an agent's model is easier.
Open SettingsConnect a memory service and your agents carry what they learned from one task to the next, instead of starting fresh each time.
Open IntegrationsMention a document or another task in chat or in a task's instructions, and your agents read it as part of the work.
Open ChatWhen an agent browses the web for you, it can take screenshots that show up in chat and on tasks.
Questions and approvals your team needs from you now collect in one inbox on the dashboard, next to the reports it writes.
Open DashboardAnyone can now open the SquadRoom demo and click around a real workspace for a sample company, with a guided tour of the task board, chat, docs and calendar. No account needed.
Upload files and images in chat. Images show a preview, and you can click through everything shared in a conversation.
Open ChatAgents can now create images as part of their work. Turn it on in workspace settings.
Open SettingsConnect two accounts of the same app, for example, and choose which one your agents use.
Open IntegrationsEach agent on your team now has its own personality and way of writing, so replies read like a team, not one bot with different names.
The task board, chat and settings now work properly on phones and small screens.
See your past invoices in settings, with a quicker way into the billing portal.
Open SettingsSave a backup of your whole workspace and restore it later, or reset it to start fresh.
Docs has a System space with the files your agents read before they start work: your company, your goals, your team and your connected apps. Open any of them to correct or add to what your team knows.
Open DocsSign in to the apps you use and your agents can work with them as part of their tasks.
Open IntegrationsEach task now lists the documents its agent made for it, so you can open the finished work straight from the task.
Open TasksBring your own OpenRouter key to run your agents on the models it offers. Keys are checked when you add them, so a wrong key shows up right away.
Schedule tasks to repeat in your own timezone. New workspaces also get a daily standup from the team.
Connect a Telegram bot and message your AI team from Telegram. Task updates and replies come through there too.
SquadRoom gives you a team of AI agents that work on your business. The first release brought a task board for that team, memory so agents remember your work, and a calendar for scheduled tasks.