An agent that runs in the cloud needs a Claude login of its own, and until now that meant copying a token by hand. Now the machine does it. You click approve once, in a browser you are already signed into, and nothing is typed or pasted. The credential is your own, issued to you, and it lives on that agent's disk rather than passing through us.
New
Hire an agent to run in the cloud instead of on your Mac, and give them a Claude login without copying anything. The machine opens an authorize URL, you approve it in a browser you are already signed into, and that is the whole interaction.
The credential is issued to you, not to us. The agent's own Claude CLI is the OAuth client: it picks the client id, the redirect and the PKCE challenge, and it performs the exchange. Dock never constructs an OAuth request, and a token minted by us would be refused at the exchange anyway.
It is written to that agent's own volume and kept at owner-only permissions, so a rebuilt machine still has the login it had before. A credential stored anywhere else works once and then quietly disappears.
It is stored exactly as the CLI wrote it, never taken apart and rebuilt, and the CLI refreshes it itself when it expires.
An agent's machine can inherit the Claude login the Mac already has, so the second agent is quicker to set up than the first.
You are asked for the login at the moment you choose cloud, not after you have finished hiring and wondered why nothing happens.
A cloud turn is refused when nothing can pay for it, and told to you as that, rather than starting and going quiet.
Polish
Choosing to fund a turn on Dock Tokens no longer changes which models you can pick.
The setup dialog names the phase it is on rather than showing an unlabelled spinner.
Hitting the cloud-agent cap now says so, instead of silently giving you a local agent.
A cancelled plan no longer keeps its cloud-agent allowance.
9 changes
Your agents get their own browser.
Some of the work only exists behind a login. LinkedIn, X, Reddit, YC: the page you need is behind a session, not an API call. Every agent now gets its own Chrome profile, signs in once, and stays signed in, separate from yours and from every other agent's. It can click, type, and fill a form, but it cannot run scripts, upload files, or reach anything on your network.
New
Agents can drive a real Chrome. Not a scraper and not an HTML fetch: the rendered page, as you would see it.
Every agent gets their own isolated Chrome profile, keyed to the agent. Their logins survive the turn, the day, and a restart.
That profile is never yours. Launching against a personal browser profile is refused outright, and the check follows symlinks so one cannot be pointed at yours.
This opens the work that needs a session: a LinkedIn search, a Reddit thread, your YC company page, a post on X. If the only way in is a login, an agent can now get there.
Agents can act, not just read: click, type, fill a form, pick from a dropdown, handle a dialog, switch tabs, and take a screenshot.
API
The browser tools are an allow-list, not a deny-list. Seventeen vetted tools are exposed and everything else is denied by default, so a new or unknown tool cannot arrive switched on.
Arbitrary JavaScript is denied: no evaluate, no unsafe code execution.
File uploads and raw network requests are denied, so the browser cannot be used to exfiltrate a file or forge a request.
Your internal network is unreachable. Loopback, private ranges, link-local, and the cloud-metadata address are blocked before the origin allow-list is even consulted, so the permissive default still cannot reach an internal service.
19 changes
GPT-5.6, Kimi K3, and GLM 5.2 join your roster.
Five new models for your teammates. GPT-5.6 Terra arrives with its heavier and lighter siblings Sol and Luna, alongside Kimi K3 from Moonshot AI and GLM 5.2 from Z.AI. Every one cleared the same bar before it got in: long real tasks, zero silent failures. Pick a model per teammate in the picker or from the CLI, and switch any time.
New
Five models join the roster: GPT-5.6 Terra, Sol, and Luna from OpenAI, Kimi K3 from Moonshot AI, and GLM 5.2 from Z.AI.
Sol and Luna are Terra's heavier and lighter siblings: the same 1M context, tuned up for the hardest work or down for the quick stuff.
Every model cleared the same bar before joining: long real tasks, zero silent failures.
Pick the model per teammate in the model picker, or dock hire --model from the CLI. Switch whenever you like.
New billing plans: Starter at $20 a month, Standard at $100, Scale at $200. When your usage runs out the choice appears in the conversation instead of a settings page, and your teammate picks the work back up once the payment clears.
Vault keys now open per agent: a new key starts locked, so nobody can use it until you add them to that key's list, and the in-chat add-key card grants the asker as you add. Keys you already had show which teammates were auto-added from earlier use, so you can trim the list.
Queued messages are a drawer: see what waits behind a busy agent, send one ahead, dismiss one, or clear the backlog. Stops land in seconds.
Polish
The model picker got cleaner: current models up front, the rest behind a More models menu.
Replies from the new models stream smoothly, with real activity shown while they work.
Agent panels open instantly and appear whole. No skeletons, no half-loads.
Table tabs switch instantly, and the row cache stays honest against your edits.
Files: folders you just visited stop re-skeletoning, and image previews open cleanly on top.
Fix
Messages in the drawer are attributed by author, so a cross-org message never reads as You.
The first-run greeting renders at the same size as every other agent message.
Delete confirmations use a real in-app modal instead of a browser prompt.
Workspace cards no longer freeze at a stale width, and docs no longer double-load on switch.
API
send_message refuses agent recipients outright instead of stranding a message silently.
request_capability_key: a tool any agent can call to summon the Vault add-key card into the conversation.
Agents can discover the files surface over MCP, with honest capability info.
29 changes
Vault is here. One shared place for your team's keys.
When a teammate needs a key mid-conversation, an add-key card appears right in the chat: a name and a value, nothing else, and they are back to work. Settings holds the full picture, and only you manage your keys. Open to every account today. The CLI grew up alongside: hire a whole roster, edit docs and tables, all from the terminal.
New
Vault, in Settings: one open surface for the keys your teammates use. Adding one is a name and a value, nothing else.
When a teammate needs a key mid-conversation, an add-key card appears right in the chat, with a proper empty state when Vault is bare.
Vault is on for every account today. No allowlist, no waiting.
dock hire creates a teammate from the terminal: name, title, role, model, manager. Or a whole roster at once with dock hire --from a JSON file.
Set a teammate's model and schedules from the CLI too. The whole lifecycle, without opening the app.
dock doc write, dock row update, and dock row rm: the CLI edits your docs and tables, with your own permissions.
Message your own agent from the web or the CLI and they wake right away. No waiting for a scheduled turn.
CLI turns get their own look in chat: clean rendering and a CLI badge. And dock agents lists each teammate's role.
A live Shipped pill on the home page rotates the three newest releases, fed by this changelog.
Polish
The home page is redrawn in the dither language: one ink, regrouped nav, app-scale buttons.
Blog, Help, and this changelog land in dark by default, and share one theme menu.
Analytics is cookieless: PostHog runs in memory and respects Do Not Track. The privacy policy says exactly that.
Workspaces open without the relayout stall, the shell is responsive, and the rail and agent pane restore where you left them.
Vault key rows read cleaner: one provider label, less noise.
Switching back to a doc is instant: the editor seeds from the prefetch cache, and the format toolbar survives workspace switches.
The rail takes arrow keys: walk every row with up and down.
Long chats render faster.
Queued messages stack in the right order, never cover the composer, and no longer jitter on send.
Onboarding: the Join-your-team list is deduped and capped, with Show all.
Fix
Vault: hardened identity checks on key pulls.
Docs no longer load twice with a double scrollbar on workspace switch.
Workspace dropdowns no longer clip, and every pill lines up.
Joining a chat mid-turn no longer shows an agent's raw reasoning.
Long replies no longer truncate at a stray control character.
Public doc links render every rich node again, not just video.
Desktop sign-in: prefers your signed-in Claude when two installs exist, keeps the keychain Allow prompt up, and a logged-out Claude gets one clear reconnect error.
Onboarding self-heals a stuck Claude connect: timeout, cancel, and an actionable fallback.
The Dock logo on Blog and Help links home again.
API
send_message answers queued_not_delivered when an agent recipient cannot take delivery, instead of pretending it sent.
7 changes
The Dock CLI. Your terminal, signed in as you.
npm i -g @trydock/cli, then dock login. The dock command acts as you from the shell, with the same permissions as the web app: message your Chief of Staff and teammates, manage workspaces, docs, and rows. Not an agent and no agent key. Your credential lives in your OS keychain, and Dock stores only a hash. Any terminal tool you already use, Claude Code included, can now drive your Dock as you.
New
Install with npm i -g @trydock/cli (v1.0.0). dock login opens the browser, you approve, done.
dock chat sends a message to your Chief of Staff or any teammate, straight from the shell.
Manage workspaces, docs, and rows from the command line, with your own permissions.
Acts as you, not as an agent. No agent key anywhere in the flow.
Your credential sits in your OS keychain and Dock stores only a hash. Revoke any time in Settings or with dock logout.
Terminal tools you already run, Claude Code included, can now drive Dock on your behalf.
API
New CLI auth flow: browser approved sign in for the terminal, revocable per credential.
8 changes
Company Brain and Second Brain for your AI teammates.
Two kinds of context, kept current without you maintaining either. Second Brain is a private picture of how you work: your goals, your voice, the people and commitments around you. Every teammate you run draws on it, so you explain yourself once. Company Brain is the shared record of how your company works: mission, who does what, decisions, priorities. Every teammate in your organization draws on that one. Your Chief of Staff starts filling both from your first conversation.
New
Second Brain: each teammate learns how you work from working with you. No forms, no questionnaire.
Second Brain is private by construction. Your colleagues cannot read it, and neither can their teammates.
Company Brain: shared company context every teammate in your organization draws on. Mission, who does what, decisions, priorities, voice.
A standing routine keeps Company Brain curated, so it does not rot the way a wiki does.
New teammates arrive already knowing the company. Day one is productive, not an interview.
Polish
The Memory tab is redesigned: everything a teammate holds, in one place.
Quick-add on the Memory tab: type what to remember, it files the note.
Tell a teammate to forget something and it is removed, not buried.
11 changes
Share your AI teammate with your whole team.
A teammate you hire no longer works only for you. Share it into your organization and colleagues chat with it directly, while it keeps one identity, one memory, and one owner: you.
New
Share an agent into your organization. Anyone on the team can open a chat with it.
Shared teammates are labeled in the agent switcher, the chat header, and the @ picker, so you always know whose agent you are talking to.
Every agent's address book lists the teammates they can actually reach, including colleagues' shared agents.
Your teammates can message each other across the organization, addressed by name.
Sign up with your company email and Dock finds your organization. Joining is automatic, with an opt out.
Create a new organization from the app. Running two companies no longer needs two accounts.
Polish
Chats with a shared teammate carry images to the recipient.
A shared teammate keeps their owner's clock. Schedules fire in the owner's timezone.
Rename a teammate and it stays reachable by its new name everywhere.
Fix
Error badges on the rail show only for agents you own, not ones shared with you.
Only agents explicitly shared with you are reachable. Nothing leaks across accounts.
10 changes
Your Chief of Staff is now on iMessage and email.
Text your Chief of Staff like a person. Email it like a colleague. Every exchange mirrors into Dock chat in both directions, so the record stays in one place. Open to all accounts today.
New
Connect iMessage from your agent's menu: scan the QR or tap to text, bound in under a minute.
Text your agent from your phone and the reply comes back as an iMessage.
Email your Chief of Staff and it answers in the same thread, with real Re: threading.
Your agent can also start an email to you when it has something worth sending.
The whole conversation mirrors into Dock chat, both directions, texts and email alike.
Locked to you: your agent sends texts and email only to its owner, and only accepts them from its owner.
Polish
Inbound texts land in Dock within seconds.
The connect panel shows exactly which phone number your agent is bound to.
An email back-and-forth folds into a single card in chat, not a pile of bubbles.
iMessage turns show a working indicator while your agent is on it.
10 changes
Workspaces are instant. Chats go five wide.
A week of speed. Switching workspaces stopped being a page load and started being a glance, and the chat row learned to hold a whole desk's worth of conversations at once.
New
Open up to five chats side by side, in one horizontally scrolling row.
Tag a workspace with @ in any chat composer. It renders as a pill your agent can follow.
A persistent workspace header with tabs, so your surfaces stop jumping around.
New workspace opens a surface picker: doc first, then table or HTML mockup.
Paste or drop a video into a doc. It plays in place.
Polish
Switching workspaces is instant: content prefetches on hover and paints in under 40 milliseconds.
Zero movement on switch. Card, header, toolbar, and column headers all hold still.
Activity, Comments, and Share redesigned as one frosted set.
Comments open inline. @agent mentions are clickable chips that open the agent's chat, and [[workspace]] cross references link straight to the workspace.
Sharing is a flat read and write list. Destructive actions stay with the owner.
8 changes
People chat is live. Message your human teammates.
Dock chat is no longer only you and your agents. DM anyone in your organization right next to your agent chats. Humans and AI teammates, one roster, one window.
New
Direct messages with any human teammate in your organization.
Emoji reactions on messages, and an emoji picker in the composer.
Click a person in the rail to open a person-only view, exactly like agents.
Polish
Consecutive messages group under one avatar, so a conversation reads like a conversation.
Share images in chat. They load with a skeleton and open in an in-app lightbox.
Attachments render above your caption, the way you expect.
Fix
Slash commands stay out of human chats. They are for agents.
Reactions and message actions are human to human only. Agent messages stay inert.
12 changes
Agent chat: streaming, queueing, and a real Stop.
The chat window caught up with what your agents actually do. You watch them think, you can stack work while they are busy, and you can stop them mid turn.
New
Live thinking: watch your agent reason while it works. On for everyone.
Message queueing: send while an agent is busy and messages hold in line. Send now jumps the queue.
Stop a running turn. It halts the work, not just the spinner.
Agents see images. Paste a screenshot straight into the composer and ask about it.
An activity line under the reply shows what the agent is doing right now.
Per agent cost dials in the composer.
Polish
Answers stream as they are written instead of dumping all at once at the end.
A busy agent no longer blocks your other teammates. Up to five turns run at once on your device.
Schedules fire in your timezone, not UTC, and every schedule shows its timezone.
Your agents inherit the connectors you already set up in Claude, with real brand icons.
Fix
Chats open at the latest message instantly. No more top to bottom scroll chase.
Renaming an agent propagates everywhere: the rail, the chat header, and the agent's own sense of who it is.
Cass's browser
