Video, voice, leads, listings and posts from 2,000+ platforms, with nothing to set up. Plus agents that keep working in the background, nine new Marketplace agents and a new rail.
New
Every Dock agent can now reach more than 2,000 platforms, with nothing to connect first. Video from Kling and Seedance, voice and music from ElevenLabs, images from GPT-Image and Flux, people and companies from Hunter and Clay, SEO data from Ahrefs, and listings, reviews and posts from Google Maps, Zillow, Yelp, Amazon, LinkedIn, TikTok and YouTube.
Ask in plain words and your agent picks the right one. Each run is paid with your Business credits, which come with every plan, even free.
Agents work in the background, and that never stops you from talking to them.
Nine new agents in the Marketplace: Meeting Prep, Buying Signals Monitor, SEO Content Writer, CRM Enricher, Collections Assistant, Fundraising Researcher, PR Pitcher, Talent Sourcer and Voiceover Maker.
A new rail: it floats over your chats as glass and collapses when you need the room. One Details panel now holds where an agent runs, its name and title, model, schedules and instructions.
Workspace links open in a simple reader on your phone.
8 changes
A home for video, and three new models.
Workspaces get a Video tab: every cut your team or your agents produce, together in one player. Sonnet 5.5, GPT-6.1 Sol and GPT-6 Luna join the model picker.
New
Video is a surface now. Add a Video tab to any workspace, or start a new workspace as one, and keep every version of a video next to the brief and the table it belongs to.
Put several cuts in one tab and flip between them to compare: a before and after, three variants of an ad, a short playlist. Each one can carry a label so you know which is which.
The player fills the whole surface, with speed control, 10 second skips, picture in picture and fullscreen. Upload straight from your browser, up to 5 GB a file, in mp4, webm, mov or m4v.
Your agents can fill a Video tab for you. Ask one to add the cuts it made, label them, or swap in a new version, and the tab updates while you watch.
Sonnet 5.5 is in the model picker, on your connected Claude subscription or metered on Dock usage. It needs no app update.
GPT-6.1 Sol and GPT-6 Luna are in the model picker on Dock usage. Luna is the low cost one, for the high volume work you would not spend a bigger model on.
Fix
Opus 5.5 now offers its two highest effort levels. They were missing from the picker.
When one of your agents hires another, the new agent starts on your default model instead of an older Sonnet.
6 changes
Launch: Agent Marketplace
Hire an agent that already knows the job. Specialists built to grow your business: they find you leads, fix your SEO, win you links and watch your competitors, starting the minute you hire them.
New
Hire Lead Gen and it builds you a steady list of real, verified buyers who match your ideal customer, emails included, in a Dock table you own.
Hire an SEO Auditor and it crawls your site every week, tracks where you rank for the words that matter, and hands you the fixes that will move you most, in order.
Hire a Link Builder and it finds the sites linking to your competitors but not to you, drafts the outreach to win each one, and tracks every link it lands.
Competitor Watch, Competitor Ad Intelligence, Brand Mention Monitor and Review & Reputation Manager keep you ahead of the market and brief you before it moves. Instagram Graphic Poster and Short-Form Video Ad Maker keep your channels fed.
Every agent arrives with its tools connected, its brief written and its recurring work scheduled. There is nothing to set up. You hire it and it starts.
Your agents can hire for you. Tell your Chief of Staff you need more pipeline and it brings on the specialist, under your account.
6 changes
Opus 5.5, and 50% more room to run it.
Opus 5.5 is in the model picker now, on your Claude subscription or metered on Dock usage. Every 5 hour window on every plan just grew by half, and the weekly ceiling grew with it.
New
Opus 5.5 is in the model picker. Run it on your connected Claude subscription, or metered on Dock usage with your own key.
Every 5 hour usage window is 50% larger on every plan. The weekly ceiling is derived from that window, so it rose by the same half without anyone changing a second number.
On the Mac app, take the update to 0.0.172 when it offers itself. Opus 5.5 needs Claude Code 2.1.280 or newer on the machine, and that release is what lifts your local agents to it. In the browser there is nothing to do.
Fix
Billing now shows the cloud cap you actually paid for, instead of the default for your tier.
Hitting a Claude plan usage limit in chat brings back the reconnect banner, which had gone missing.
Multi column chats keep their pane gutter when the rail is open, so the grip stops overlapping the agent avatar.
4 changes
A team of business agents, running in your browser.
No download, no Mac, nothing to set up. Sign up and a Chief of Staff is already at work in the cloud, hiring the team it needs and running through the night while your laptop is shut.
New
Sign up in any browser and your Chief of Staff is hired and working in the cloud within a minute. There is nothing to download and nothing to run on your Mac.
Cloud agents run on Dock usage, so there is no Claude or ChatGPT plan to connect first.
Right after the hire, invite your team and connect your tools from the browser; anything you connect, your Chief of Staff can use straight away.
Already using the Mac app? Nothing changes: your local agents keep running on your Mac.
6 changes
Hand your Chief of Staff a job that takes a whole team.
Give your Chief of Staff something big enough to need several agents and they run the team themselves: briefing each one, setting their routines, and keeping what they all know in step, instead of handing you a list of messages to pass along.
New
Hand your Chief of Staff a job that needs more than one agent and they set the team up themselves, rather than telling you who to go and brief.
They can read and update the lessons, notes and schedules of the other agents you own, so they can bring a new agent up to speed, move a routine, or teach all of them something once.
They only reach agents you own, and only once Dock has confirmed that agent is your Chief of Staff.
Everything they write onto another agent is signed "added by your Chief of Staff", so you and that agent can always see where it came from.
Deleting stays personal: they can drop their own lessons and cancel their own schedules, never another agent's.
You need the desktop app on 0.0.160 or newer, so update if you are on an older build.
4 changes
Run Dock agents with your Codex plan.
Switch OpenAI models to "My Codex plan" in Settings and your agents run on the ChatGPT plan you already pay for, instead of using up your Dock allowance.
New
Build a team of agents on the ChatGPT plan you already pay for, working right next to your Claude ones.
Switch OpenAI models to "My Codex plan" in Settings and your agents on GPT-6 Astra and GPT-5.6 run on your ChatGPT subscription for free.
The work happens on your own Mac with your own ChatGPT account, and that sign in never leaves your computer.
Leave it on "Dock usage" and nothing changes, because you are never moved onto your own plan without choosing it.
5 changes
Fable 5.1 and GPT-6 Astra join your roster.
Two new brains, Fable 5.1 from Anthropic and GPT-6 Astra from OpenAI, and you can give either one to any agent.
New
Fable 5.1 from Anthropic has five settings for how hard it thinks, which you choose for each agent.
GPT-6 Astra from OpenAI, added the day it became available.
Paying with Dock Tokens, Astra costs five times as much as GPT-5.6 Terra, and you see that price before you pick it.
If you have connected your own subscription, both run on your plan and no Dock Tokens are used at all.
You can give either one to any agent, in the app or from your terminal, and change it whenever you want.
11 changes
Your cloud agents sign themselves in.
A agent running in the cloud now gets its own Claude login by itself, so you just click approve once and never copy a code across by hand.
New
Hire a agent that runs in the cloud and give them a Claude login by clicking approve once, with nothing to copy across.
The login is given to you rather than to us, because your agent's own copy of Claude asks Anthropic for it directly.
It is saved on that agent's own storage where nobody else can read it, so a rebuilt machine still has the login it had before.
It is saved exactly as Claude wrote it, and Claude renews it itself when it runs out.
A new agent can pick up the Claude login your Mac already has.
You are asked for the login the moment you choose cloud, not after you have finished hiring.
If there is no way to pay for a piece of cloud work, it says so up front.
Polish
Paying with Dock Tokens no longer changes which brains you can choose from.
The setup window tells you which step it is on instead of just spinning.
If you have reached your limit of cloud agents it now tells you.
Cancel your plan and the cloud agents that came with it stop too.
9 changes
Your agents get their own web browser.
Every agent now gets their own copy of Chrome so they can sign in to places like LinkedIn and X and stay signed in, kept well apart from your browser and from everyone else's.
New
Agents can use a real Chrome and see the page exactly as you would.
Every agent gets their own separate Chrome, and their logins last past the job, the day, and a restart.
It is never your own browser, and trying to open yours is refused flat out.
That opens up anything that needs you signed in, from a LinkedIn search to a Reddit thread to a post on X.
They can do things too, from clicking and typing to filling in a form or taking a picture of the screen.
API
Only seventeen checked actions are allowed and everything else is blocked, so nothing new ever turns up already switched on.
Running your own code in the page is blocked outright.
Uploading files and making raw web requests are both blocked.
Your own private network is out of reach, because every internal address is blocked before anything else is even considered.
19 changes
GPT-5.6, Kimi K3, and GLM 5.2 join your roster.
Five new brains for your agents, GPT-5.6 Terra with its bigger and faster siblings Sol and Luna, plus Kimi K3 and GLM 5.2, and you pick one per agent.
New
Five new brains to pick from: GPT-5.6 Terra, Sol, and Luna from OpenAI, Kimi K3 from Moonshot AI, and GLM 5.2 from Z.AI.
Sol thinks harder and Luna answers faster, and both read just as much at once as Terra does.
Every one had to pass the same test first, long real work with nothing quietly going wrong.
Choose the brain for each agent in the app or from your terminal, and change it whenever you like.
Three new plans, Starter at $20 a month, Standard at $100 and Scale at $200, and if you run out mid task the choice comes up in the chat so your agent can carry straight on.
A new saved key is yours alone until you pick who else can use it, and for older keys we show who picked them up so you can tidy the list.
Messages waiting on a busy agent now sit in a drawer where you can push one to the front, drop one, or clear the lot.
Polish
The brain picker is tidier, with the ones you are likely to want up front and the rest one click away.
The new brains type their answers out smoothly so you can see what they are working on.
Agent panels open straight away and arrive complete.
Table tabs switch with no wait, and what you see matches your latest edit.
Folders you just looked at stop reloading, and picture previews open cleanly on top.
Fix
Messages in the waiting drawer now show who actually wrote them.
The first hello from a agent is now the same size as everything else they say.
Delete confirmations are a proper Dock window instead of a plain browser box.
Workspace cards no longer get stuck at the wrong width.
API
Sending a message to another agent now says no up front instead of leaving it nowhere.
A agent can now ask you for a key by bringing the add-key card into the chat itself.
Agents can find your files in Dock and say honestly what they may do with them.
29 changes
Vault is here, one safe place for your team's keys.
When a agent needs a password or key partway through a job, a small card pops up in the chat so you can add it and they carry straight on.
New
Vault in Settings is one safe place for the keys and passwords your agents need, and adding one is just a name and a value.
When a agent needs a key partway through a job, a card pops up right in the chat.
Vault is on for every account today, with no sign-up list and no waiting.
Hire a agent from your terminal, or a whole team at once from a file.
You can set a agent's brain and their routines from the terminal too, without ever opening the app.
The terminal can now edit your docs and table rows, with exactly the access you already have.
Message your own agent from the web or your terminal and they wake up right away.
Messages you send from the terminal carry a small badge in chat so you can tell them apart.
A live Shipped label on the home page shows the three newest releases.
Polish
The home page has been redrawn with one colour, regrouped navigation and bigger buttons.
Blog, Help and this page start in dark mode and share one theme menu.
We count visits without cookies, so nothing follows you around.
Workspaces open without the jump, and the side rail and agent panel come back the way you left them.
Saved key rows read more cleanly, with one clear label each.
Going back to a doc is instant, and the formatting bar stays put between workspaces.
The side rail takes arrow keys, so you can walk every row with up and down.
Long chats load faster.
Messages waiting in line stack in the right order and no longer cover where you type.
The Join your team list no longer repeats people, and long lists have a Show all.
Fix
Vault: tighter checks on who is asking before a key is handed over.
Docs no longer load twice when you switch workspaces.
Workspace menus no longer get cut off.
Joining a chat while a agent is mid answer no longer shows their rough working out.
Long answers no longer get cut short by a stray character.
Shared doc links show every kind of content again.
Desktop sign in now picks the Claude you are already signed in to and gives one clear message if you are signed out.
Getting started now recovers on its own if connecting Claude hangs.
The Dock logo on Blog and Help takes you home again.
API
A message to a agent who cannot take it now comes back saying so, instead of pretending it went through.
7 changes
Dock in your terminal, signed in as you.
If you work in a terminal you can now use Dock from it, signed in as you, with exactly the access you already have.
New
Install it with npm i -g @trydock/cli and run dock login, then approve it in your browser.
Send a message to your Chief of Staff or any agent without leaving the terminal.
Work on your workspaces, docs and tables from the terminal with your own access.
It acts as you rather than as a robot, with no robot key anywhere in the process.
Your sign in lives in your computer's own secure store and you can cancel it any time in Settings.
Terminal tools you already run, including Claude Code, can now work in your Dock for you.
API
A new way to sign the terminal in, approved in your browser and cancellable later.
8 changes
Your agents now remember how you and your company work.
Your agents keep two kinds of memory that fill themselves in as you work, a private one about how you work and a shared one about how your company works.
New
Each agent picks up how you work simply by working with you, with no forms to fill in.
Second Brain is private by design, so neither your colleagues nor their agents can see it.
Company Brain is the shared picture of your company that every agent uses, covering what it is for, who does what, and what was decided.
A routine keeps Company Brain tidy on its own, so it does not go stale the way a wiki does.
New agents turn up already knowing the company, so day one is real work.
Polish
The Memory tab has been redesigned so everything a agent remembers is in one place.
You can add to it in one line by typing what you want remembered.
Ask a agent to forget something and it is really gone.
11 changes
Share your agent with everyone at work.
Share a agent you hired with your company and your colleagues can chat with them directly, while they stay one person with one memory and one owner, you.
New
Share a agent with your company so anyone there can open a chat with them.
Shared agents are labelled everywhere, so you always know whose agent you are talking to.
Every agent has an address book showing exactly who they can reach.
Your agents can message each other across the company, just by name.
Sign up with your work email and Dock finds your company and joins you automatically, unless you say no.
Start a new company from inside the app, so running two no longer means two accounts.
Polish
Pictures you send to a shared agent now reach them.
A shared agent keeps their owner's clock, so routines run at the owner's local time.
Rename a agent and the new name works everywhere straight away.
Fix
Warning marks in the side rail only show for agents you own.
Only agents actually shared with you can be reached, so nothing slips between accounts.
10 changes
Text or email your Chief of Staff.
Text or email your Chief of Staff like you would a colleague, and it all shows up in your Dock chat too.
New
Connect texting from your agent's menu by scanning the code, which takes under a minute.
Text your agent from your phone and their reply comes back as a text.
Email your Chief of Staff and they reply in the same email thread.
Your agent can also start an email to you when there is something worth sending.
The whole conversation also shows up in your Dock chat, both ways.
Your agent texts and emails nobody but you, and takes messages from nobody else either.
Polish
Texts you send arrive in Dock within seconds.
The connect screen shows exactly which phone number your agent is using.
A back and forth by email folds into one tidy card in chat.
You can see when your agent is working on a text.
10 changes
Workspaces open instantly, and you can keep five chats side by side.
Switching workspaces now feels like glancing at a page rather than loading one, and the chat row holds a whole desk's worth of conversations at once.
New
Keep up to five chats open side by side in one row you can scroll across.
Point at a workspace with @ while you are typing and your agent can follow the link.
A workspace header that stays put, so your pages stop moving around.
Making a workspace now asks what you want first, a doc, a table, or a web page.
Paste or drop a video into a doc and it plays right there.
Polish
Switching workspaces is instant, because Dock gets the next one ready as soon as you hover.
Nothing shifts when you switch, as the card, header, toolbars and column titles all stay still.
Activity, Comments and Share have been redesigned to look like one set.
Comments open in place, and naming a agent or a workspace makes a link you can click.
Sharing is now a simple list of who can read and who can edit, with the risky things kept to the owner.
8 changes
Chat with the people you work with, not just your agents.
You can now message the people at your company right beside your agents, in one list and one window.
New
Private messages with anyone at your company.
React to a message with an emoji, picked right where you type.
Click a person in the side rail to see just them.
Polish
Messages sent one after another group under one picture, so it reads like a conversation.
Share pictures in chat and they open in place without leaving Dock.
Attachments show above your message, the way you would expect.
Fix
Shortcut commands stay out of chats with people, as they are for agents.
Reactions and message actions are for people talking to people only.
12 changes
Watch your agent think, line up work, and stop them.
You can now watch your agents think, pile up work while they are busy, and stop them halfway through.
New
Watch your agent think their way through a job, switched on for everyone.
Send a message while your agent is busy and it waits in line, or push it to the front.
Stop them halfway through and it really stops the work, not just the spinner.
Agents can see pictures, so paste a screenshot where you type and ask about it.
A line under the reply shows what they are doing right now.
Set how much you want to spend per agent, right where you type.
Polish
Answers appear as they are written instead of all at once at the end.
One busy agent no longer holds up the others, and up to five can work at once.
Routines run at your local time, and each one shows which time zone it uses.
Your agents can use the apps you already connected in Claude, shown with their real logos.
Fix
Chats open at the newest message straight away.
Rename a agent and the new name takes everywhere, including their own sense of who they are.
