Claude Interactive Connectors: Software Now Lives Inside Claude

Claude can now open the full version of a tool inside the chat. If you build or sell software, that changes where your product lives, and it is worth planning for now.

On 10 November 2026, Anthropic announced interactive connectors: you can work in Adobe, Atlassian and Box beside your conversation, and make changes yourself or ask Claude to make them with you. It is rolling out on web and desktop.

What Claude Interactive Connectors are

Interactive connectors are not brand new. Earlier this year, Anthropic launched them to bring a small live view of other tools into Claude. What is new is that now, where a connector supports it, you can open the whole tool beside your conversation, or expand it to full screen.

On its announcement page, Anthropic explains what that means in practice:

• Claude works in your open tools. You ask it to change a page, a row or a file you have open, and you approve its actions the same way you already approve a connector's actions. The changes appear in the tool as Claude makes them.
• You get the full editor. For tools like Confluence or Notion, you see the real editor, signed into your own account, not a limited preview.
• No prompt needed to open it. You can open a connected tool directly and ask Claude for help from the same screen.
• Pin the ones you use most. Tools can be pinned to the sidebar and opened with one click.

In the frame, a Box file open beside the chat with Claude (source).

Anthropic's examples are concrete. With Adobe, you ask Claude to retouch and edit photos, or to highlight, edit and redact PDFs. With Atlassian, you open a Confluence page beside the chat, ask Claude to write and edit, and watch the changes land on the page. With Box, you open a file, ask for a revision, and compare the two versions before you keep the changes.

Beyond those, Anthropic says it is starting with a small group of tools from its directory, including Asana, Canva, Carta, Dropbox, monday.com, Smartsheet, Vercel and Webflow, plus HubSpot and Notion. Figma, Harvey and Shopify are building their own integrations.

It is rolling out on web and desktop, on all plans, as part of the new Claude experience that brings chat and Cowork together. You see it once your account is updated or your organization turns it on, so it does not switch on for everyone at the same moment.

Our take: software is becoming Claude-centric

The center is moving: software is becoming Claude-centric. And that is much bigger than a new button in a chat.

For years, the habit was simple. You opened the app, and the app was the workplace. If you wanted help from an assistant, you copied something out, asked a question and pasted the answer back. The app sat in the middle. The assistant sat on the side.

Interactive connectors flip that layout. The conversation becomes the workplace, and the app opens as a window inside it. You stay in Claude, and Adobe, Atlassian or Box comes to you.

More and more daily work will start in the assistant, and the product will show up when someone needs to see or change the real thing: a page, a file, a photo, a record. The interface still has to be clear, but it matters less as the place people spend their day. What matters more is the data the tool holds and the actions a person or Claude can take on it, with an approval in between.

That does not mean standalone apps disappear tomorrow. People will still open the full product when they need room, or when they are not in Claude. It means the default path is starting to change.

Data is moving into Claude too

In the same thread from @claudeai on X, Anthropic took another step in the same direction.

Claude Data, formerly Dashboards, is now generally available on paid plans. Along with dashboards, Claude can draft data stories and narrated videos from your metrics, for you to review before sharing. Certified queries and live sharing are in beta.

The same day, Anthropic also started rolling out Claude Whiteboards and Claude Sheets, in beta on paid plans. No need for a full tour here: we mention them because they point to the same idea. More of the work stays inside Claude, instead of being copied out to another tab.

What changes if you use software

If you spend your day in tools like Confluence or Box, the change is mostly a habit. Instead of opening the app and hunting for the right feature, you start from the request: "draft the launch plan", "review this document", "fix this photo". Then you look at the result in the tool open beside you and correct it yourself, or ask Claude to.

Two tips for when it reaches your account:

• Start with a task you repeat often. A document you write every week, or a check you always do the same way. That is the easiest place to see whether it actually saves you time.
• Always review before you keep it. That is the whole point of having the tool open next to the chat: seeing with your own eyes what changed before you approve it.

What changes if you build software

If you are a founder, build a product or sell software, the useful question is not "is this a nice demo?". It is: if a customer can do the work that matters beside a Claude chat, what is your product for?

The answer has two parts. First, your product is still where the data lives. Claude does not replace the file in Box, the page in Confluence or the photo in Adobe: the window inside Claude is just a way to reach them. Second, what the customer feels is the action, not the tour of your interface. Can they make the change? Can they see it happen? Can they approve it? Can they compare versions before they keep it?

Five things to do this week, even before it reaches your account:

1. List the objects and the actions, not the screens. Write down the real objects your product holds, like a page, a file, a record or a task, and next to each one the actions that matter: create, edit, review, compare, approve, share. That is what would travel into a window inside Claude.
2. Design for work beside a conversation. When your tool opens next to a chat, the customer should see the object, the change, and a way to approve or reject it.
3. Put review in the path, not after it. Claude Data drafts stories for you to review before sharing, and with Box you compare versions before you keep a change. If your product changes real data, the review step is part of the product.
4. If you build integrations, look at MCP Apps. Anthropic says interactive connectors run on MCP Apps, an open extension to the Model Context Protocol. An app built this way is not tied to Claude: it runs in any AI application that supports the extension. The newer capabilities, such as a full embedded app, are in public preview for developers.
5. Write a product sentence that still works inside a chat. "You can revise the file and compare versions before you keep the change" works in a window beside Claude. "Here is our beautiful workspace" does not.

For websites, the idea is the same one we cover in our piece on AI-friendly sites with WebMCP: make it clear to an agent what it can read and what it can do. What to sell when AI can rebuild an app in an afternoon is a different question, and it gets its own piece.

Keep in mind

• The rollout is gradual. The features are the ones Anthropic describes on its announcement page: check what you see on your own account and when.
• Not every connector supports it. The full view is there where the connector supports it. Before you change how you work, check that the tools you actually use are included.
• Some parts are in beta. Whiteboards, Sheets, and certified queries and live sharing in Claude Data may still change.

Want to work out how to get your product, or the way your team works, ready for this shift? Get in touch and let's talk it through.

In short: Software is moving inside the assistant: the app becomes a window beside the conversation. If you build a product, get the data and the actions right first, and the screens second.