Claude can now open Adobe, Atlassian and Box right inside the chat

A familiar scene: you ask an AI assistant to sort out a task in a tracker or prepare a document, it produces a tidy answer, and then the manual work begins. You copy the text, switch to the tab with the right service, paste it, edit it, and go back to the chat with a follow-up. Every such jump eats context and time. Anthropic has announced that it intends to fix this.

What exactly was announced

The post from the @claudeai account says that with interactive connectors, the full version of a tool can be opened inside Claude. Adobe, Atlassian and Box are named as examples. You are meant to work with them alongside the conversation. You can change things yourself, or you can ask Claude to make the edits together with you. The rollout is happening on the web and desktop.

This is an important distinction. An ordinary connector in chat assistants is primarily an access channel: the model reads data from a service and sometimes does something in it. You see the result in the text of the reply. An interactive connector, judging by the wording, adds a second layer: the service's actual interface lives in the same window as the conversation.

What this might look like in practice

The announcement contains no detailed scenarios, so what follows are logical examples rather than confirmed features.

A product manager discusses a release plan with Claude. A task board from Atlassian is open alongside. They move one card themselves and ask the model to rephrase the next ten tickets. A designer or marketer opens a file in Adobe and explains in the chat what needs to change, making a small tweak by hand. A team whose documents are stored in Box views a file next to the conversation and immediately asks Claude to produce a summary or flag contentious passages.

The key idea is that the person and the model work on the same object rather than handing copies back and forth. Fewer discrepancies between versions, fewer instances of "are you sure you pasted the latest revision?"

Why this is worth noticing

The chat interface has long been a bottleneck: it's convenient for asking, inconvenient for finishing the job. If the assistant turns into a work surface that hosts third-party applications, competition changes too. The winner is not the one whose model is a couple of points better on a benchmark, but the one in whose window the user spends the working day.

For product managers and founders building SaaS, this is a signal of two kinds. First, your product may end up not as a separate tab but as a panel embedded inside someone else's assistant. Second, the value of an interface is now determined also by how well a model can work with it, not just a human.

What we don't know yet

Several things do not follow from the post, and they are worth checking before making plans:

• Which plans get the feature and when the rollout will reach all users.
• Whether third-party developers can plug in their own tools in this format, or whether the partner list is closed.
• How permissions work: what exactly Claude can change in a service without confirmation.
• How complete the "full version" of a tool really is, and whether there are cut-down scenarios.
• Whether the same will appear in mobile apps: for now only web and desktop are stated.

The last point is especially important for teamwork. When a model can change real documents and tasks, the question of auditing edits and rolling them back becomes practical rather than theoretical.

What to do right now

If your team already uses Adobe, Atlassian or Box, it makes sense to check whether the feature is available in your account and try it on one non-critical workflow. Measure a simple thing: how many tab switches and text copy-pastes a typical task takes before and after.

If you build a product with its own interface, watch whether Anthropic opens this format to third-party developers. For now it is one post and three named partners, but the direction is clear: the assistant wants to be not a conversation partner but a desktop.