What's new

New

• When the agent needs a connector your Claude account doesn't have yet, like Google Docs or Slack, the card says so with one button to claude.ai's Connectors page.
• Come back to Slip after turning it on and the card reconnects by itself, keeping the conversation. No restarting the card.
• Settings › Connections lists the connectors from your Claude account, with a button to connect more.

New

• In an iOS project, the phone button opens the Simulator on the device your app is running on, where you can scroll and tap. If nothing's running, it builds and runs.
• The arrow beside it picks the app, the device and the scheme. Pick an iPad and the running app moves there.
• Click through to a screen in the Simulator, then press "screen" by the message box to add exactly what you're looking at to your message.

New

• Your skills work in every card. Claude, Codex and Grok can all use the skills on your Mac, plugin skills included. Type / in any card, or just ask.
• Settings › Skills is one list with a filter: yours, your project's, your Claude account's, plugins' and Claude Code's own.
• Make a Skill… opens a card that writes a new skill with you, with a few examples to start from.
• When an agent uses a skill, the conversation says so: "✦ using the gsap skill", the same in every agent.
• The iOS preview finds apps inside a bigger project (like apps/your-app), and shows the app the agent built and launched itself, without building it again. With several apps, pick one from the preview.

New

• The card header says where your work stands. Green "up to date" means nothing new on GitHub and nothing of yours waiting. A dot with a count (● 1) is work still only on this Mac, and ↑2 is saves not yet on GitHub.
• Click either to see the files, with how many lines changed, or the saves that haven't gone up yet. [push to GitHub] is right there, and if teammates pushed in the meantime, it brings their work in first.
• Type / and the skill you meant comes first: names that start with what you typed, then plugin names, then skills whose description mentions it.

New

• "What does this do?" now sits beside the command, not in the list of answers. The plain-words explanation opens right under the command it describes.

New

• Permission prompts lead with what the agent is trying to do, and say why Slip is asking in plain words. Not sure? Pick "What does this do?" and Slip reads the command, and the script it runs, then explains it right in the prompt. The agent keeps waiting, so nothing stops.
• Type / and every skill says what it's for, in its own words. Letters match anywhere in the name, like in the terminal, and /skills lists them all.
• Settings has a Skills page: yours, each project's, your Claude account's, your plugins' and the ones built into Claude Code. Make a Skill… opens a card that writes a new one with you.

New

• Opened Slip straight from Downloads? It says so and offers [move to Applications]. One click copies it there, lets your cards write their briefs, reopens from Applications and puts the leftover copy in the Trash. On a Mac where you can't write to Applications, it uses the Applications folder in your home folder instead.

New

• Control-C stops a running turn, like it does in a terminal. Esc and ⌘. still work too.

New

• Type / and your Claude Code skills are listed after Slip's commands: your own, the project's, and your plugins'. Pick one and it goes in the box for you to finish.

New

• A card can start a new card for the next piece of work. The agent writes the new card's brief from your conversation and asks "Start a card to…?". Say start it, and the new card opens in its folder, already knowing the goal and the next step. The card you were in stays put.
• A card can also offer to move itself to another project, bringing what you decided and what's next with it.
• Start a card from here… in the project menu begins that ask for you.

New

• The card header shows where the project lives on GitHub, like github/owner/repo. One click opens your review, your branch or the repo, and hovering says whether your work is on GitHub yet.
• Click the project's name in the header for its folder, Show in Finder, Open on GitHub, and switching branches. Changing the project is still there, one step further in.
• The card's name lives in the window title, where you rename it. The header row keeps to the project.
• The app and design icons sit on the right, beside the context meter. The design icon is now a pen nib.
• Closing a card shows "writing brief…" on its project's tile, so you can see it packing up even from the main view.
• The Pro sheet is simpler: two plans side by side at the same height, shorter lists, and one clear Get Pro button.

New

• The card header says what the card is: project, part and name. Beyond that it stays quiet, and speaks up only when something changed out there: "3 new [update]" when teammates push, with who and what on hover; "in review" and "live" once your work is out; and a plain "didn't finish" or "didn't go live" with [see why] when something needs you.
• Cards name themselves from your first message. Click the name in the header, or the window title, to rename one.
• The preview is two icons beside the card's name: the app, with a dot while it's running, and the design.
• Opening the design with nothing to show offers to sign in to Figma right in the pane, then asks which file is this card's design.
• Type /plan followed by what you want and the card switches to plan mode and sends it. /ask, /auto and /full work the same way.
• When a new server starts, including one the agent starts in its own copy of the project, the preview follows it.
• Codex and Grok cards that need a sign-in say so, and one click signs in and starts the card. The model menu marks agents that need signing in and folds older models under "previous models".
• Compare is gone. The app and the design sit one click apart in the header instead.

New

• Cards gather into projects on the home screen, one tile per repo. Click a project to see its cards, move a card into another with Move to or by dragging it, rename a project, and give it its own colour.
• Archived cards have a page of their own, grouped by project, so you can tell what each one was. Restore one, or a whole project at once, and archive a finished project in one step.
• When Free's three cards are full, the sheet lists them with [archive] beside each, and archiving one goes straight on to the card you were making.
• When the agent asks a multiple-choice question, you can answer "none of these" or type your own.
• The preview has three views: app, design and compare. Design shows the frames the agent fetched from Figma or Paper at full size, and [send this frame] puts one in your message. Compare lays the frame over the running page at the same width, as a fade or a line you drag.
• A card remembers its Figma file. Paste a link once and the header names the file, and the agent knows which file "the design" means from then on.
• [update] brings in the team's newest work and keeps your edits in progress, and new branches start from the team's latest.
• Quitting Slip lets open cards write their briefs first. Press ⌘Q twice to quit straight away.

New

• A card you're done with is archived, not deleted. It leaves the grid for a shelf underneath, with its brief, decisions and conversation intact, and comes back with one click. Archiving frees a Free slot; deleting is still there on the shelf, behind a confirmation.
• Pro is paid from the start. Free is three cards with no account; Pro is $7 a month billed annually, or $12 month to month, and starts when you buy it. There's no sign-in trial any more.

New

• Free is three cards — any projects, all running at once. A fourth card is where Pro starts, and that's where Slip asks: when you're starting more work, never in the middle of it. Delete a card you're done with and its slot comes back.
• No account needed for Free. The setup's account step can be skipped; sign in when you go Pro.

New

• Claude cards show usage against your plan beside the last turn's token count — "Max 6% of 5h · 7% of week" — with reset times on hover and amber past 80%. It's the same figure Claude Code's own /usage shows. The Account pane's Claude row carries it too. Codex and Grok don't report plan limits, and the token line says so.
• Drop any file onto a card — a .md, a PDF, a folder — and its path goes into your message for the agent to read, the way an @ mention does. Pictures still attach.

New

• Drag the divider beside the preview and the page follows the pointer; let go near a common width — 390, 768, 1024, 1280 — and it lands there. The pane's footer shows the width you're seeing the page at, with phone, tablet, laptop and desktop a click away.
• The conversation gives way to the preview, down to 280 wide, so a 1440 window can show a page at 1024.
• Pro is $7 a month billed annually ($84 a year). Month to month stays $12.

New

• A new card can start from a GitHub link: paste the project's address and Slip puts a copy on this Mac and opens the card on it. The same link pasted into a card with no project does the same — or attaches the copy that's already here.
• A card's project can be added later, too: "pick a project" in the header asks whether it's on this Mac or on GitHub.
• /plan works on every agent. Claude enters plan mode; Codex and Grok are asked to look and propose before changing anything, with their asking mode still guarding writes.
• A picture the agent read opens as large as the screen allows, with buttons to open the original full size or save a copy.
• The empty card's suggestions fill the message box for you to edit instead of sending on their own.

New

• Send feedback from the home window: one note, screenshots pasted or dropped in, and the build details ride along so a reply never starts with "which version?".
• When an agent can't answer — its included usage is used up, a spending limit is hit, it's rate-limited, or it's having trouble on its end — the card says so in plain words and offers the way out: continue with another agent, try again, or upgrade.
• Claude Code stays current on its own. The model names in the menu — fable, opus, sonnet — always mean the newest of that tier, and Slip updates Claude Code in the background when no card is connected, then tells you what opus means now. A switch in Defaults turns it off.
• The model menus say what each name means today: "opus · Opus 5.5".
• ⇧⏎ inserts a newline in the message box.

New

• A card can work in its own copy of the project: pick "new branch…" from the branch name and the card moves there, edits in hand, while the other cards' files stay put.
• [share] and [ship] say what they're about to do first, and the second click is the go-ahead.
• A project that isn't on GitHub yet gets [save] instead of [share], and a clean tree with nowhere to share shows no verb at all.
• Xcode 27 support: the simulator pane runs the device headlessly and shows its screen, since Xcode 27 has no Simulator app.

New

• Settings has one Account pane: you at the top, and the agents signed in on this Mac beneath.
• Integrations is now just the tools your agents can reach: dev tools, design tools, MCP servers.

New

• A card waiting on your answer says so and stops the clock, so an overnight permission prompt reads "waiting for you" rather than seventeen hours of "working".
• Long turns read as "4m 12s" or "1h 03m", and the "usually here" baseline only counts time the agent was working.
• The header tells you what stage your change is at — draft, saved, shared, live — and offers the one next step.
• A card with no project asks what you want to do: an existing project by name, a new one, or a question.
• Two cards on the same project know what the other is doing, and say so when both edit the same file.
• Click [preview] on a web project and Slip starts the dev server for you.
• The connectors line names the connector, says it's optional, and stays dismissed once you close it.
• "Pick up where we left off" shows as the button you pressed, with the exact prompt one click away.

New

• Setup ends by opening your first card. It asks what you're working on and whether it lives on this Mac, somewhere else, or nowhere yet, then walks you through the two or three things in Slip that matter for it.
• Your brief sits above the conversation and fills in as you talk. It's the file the card carries forward, and you can edit it any time.
• The walk-through covers the tools you've actually connected, not a catalogue.
• Founding status shows right after sign-in.

New

• In the preview, click the scope icon, then click any element on the page, and say what should change. The agent gets your note with the element named and a picture of it.
• Setup asks for your email right after Claude connects. A six-digit code, no password. Everyone has an account now, free or paid, so a purchase always has somewhere to land.
• Bought Pro on the site before opening the app? Sign in with the same email and the plan is yours on the spot.
• Setup got the site's look: the icon, the wordmark, and the glows behind a frosted card. Rows are bigger, and continue is highlighted when nothing else needs a click.

New

• Sign in with your email from Settings › Slip Account. A six-digit code arrives in seconds; no password, no link to click.
• Signing in starts a 14-day Pro trial. Sign in during the beta and you're a founding member with six months of Pro.
• Slip Pro: unlimited agents running at once, $9 a month billed annually or $12 month to month. Checkout is on slip.design; the app never sees a card.
• Manage billing from Settings, straight into the customer portal: invoices, card, plan, cancel.
• Drag across a reply and the selection runs through bullets and paragraphs alike, so copying a whole answer is one motion.

New

• Updates download and verify in the background, so the strip says “ready to install” and [update now] is just the swap.
• Before an update restarts Slip, every open card writes its brief, the same way closing the window would. You come back to briefs, not recoveries.
• Install any release from the feed, not only the newest.
• A release marked critical shows an amber strip with no [later].

New

• Scroll up in a transcript and it stays where you are. It follows new output only while you're at the end.
• Paths the agent prints open in Finder.
• Questions can fold down to a strip while you keep working.
• Several connectors sign in from Settings, and the Claude account strip comes first in setup.

New

• Slip updates itself. Every update is checked three ways before it runs: the feed signature, the code signature, and that it comes from the same developer as the app you're running.
• Auto mode is stricter after an independent security review. Compound shell commands, branch deletion, edits outside the card's folder, and a dozen other things now stop and ask you.
• Downloads come from GitHub Releases.

New

• Settings shows the name of each Claude model (“Fable 5.1”), learned from the CLI you have installed.
• The model picker becomes a menu when an agent offers more than four models.

New

• Codex and Grok are detected at launch, so /model offers them without a visit to Settings.

New

• A card can run on OpenAI Codex or xAI Grok Build instead of Claude Code. Each agent signs in through its own CLI; Slip never holds a key.
• Pick another agent's model under /model and the card moves there, brief carried over.
• In-app updates, starting with this release.