Claude can now read and edit your Google Docs, Sheets and Slides, as long as you are on a paid plan

Anthropic walks Claude into Google's house

Most AI assistants ask you to come to them. You open a chat window, paste in your work, read the reply, and copy it back. On 6 October 2026 Anthropic tried to close that loop inside the place a lot of people actually do their work, announcing Claude for Google Workspace: a public beta that puts Claude directly into Google Docs, Sheets and Slides.

There are two doors into it, and they point in opposite directions. The first is an add-on. "With the add-on, Claude can read the doc, sheet, or deck you have open, see which text, cells, or slides you've selected, and make changes directly in the file," Anthropic says. In practice Claude opens in a sidebar next to whatever you have on screen, so you are not switching between a browser tab and a chat app while you write or build a spreadsheet.

The second door is a set of connectors, also in beta, that work the other way. You start in Claude, paste a link to a compatible Google file or simply ask for a new one, and Claude creates or edits the Doc, Sheet or Slides deck from the chat. The add-on brings Claude to your file; the connectors let Claude reach out to your Google Drive from its own window.

The paywall and the permission wall

Two limits define who this is really for. The first is money. Claude for Google Workspace is available on all paid Claude plans, meaning Pro, Max, Team and Enterprise, and not on the Free plan. Anthropic did not attach a separate price, so for paying users it is included, but anyone on the free tier is shut out. This is a feature built to make paid subscriptions stickier, not to widen the funnel.

The second limit is the one that should reassure IT departments. Anthropic says Claude's access matches your existing Google sharing permissions. Claude does not get a master key to your Drive; it can reach only what your own Google account can already reach. For companies, the enterprise controls Anthropic already offers carry over to the add-on, including the Compliance API, customer-managed encryption keys, and OpenTelemetry audit export. None of that is new technology, but extending it to an assistant that edits live documents is the part a security team will ask about first.

That permission model is the quiet throughline across Anthropic's recent product moves. The same instinct to keep Claude inside bounded, auditable surfaces shows up in its push into dashboards and other first-party file types: give the assistant more to do, but fence off what it can touch.

Why this is a harder fight than it looks

The feature is clean. The strategic position is not. Google Workspace is Google's product, and Google's own assistant, Gemini, is already built into Docs, Sheets and Slides natively. Anthropic is arriving as an add-on and a set of connectors layered on top of a suite whose maker ships a competing assistant for free to the same users.

That changes what "Claude works where you work" has to mean to win. It is not enough to be present in the file; Claude has to be noticeably better at the editing task than the assistant Google already put one click away. And on that question, the announcement is silent. Anthropic describes what Claude can do inside Workspace in detail, but publishes no benchmark, no evaluation, and no comparison of how well Claude edits a spreadsheet or redrafts a slide against Gemini doing the same job. The capability is documented; the quality is a claim you have to test yourself.

There is a real argument for Claude here, and it is about the model rather than the integration. Teams that have standardized on Claude, often on the back of models like Sonnet 5.5, may simply prefer its writing and reasoning and want that specific assistant in their documents rather than whichever one their office suite happens to bundle. For those users, being able to stay in Claude and still touch their Google files removes a daily annoyance. For everyone else, the native option is already there and already free.

What to watch

Two things will decide whether this matters. The first is edit reliability. An assistant that reads your selection and changes the file in place is useful only if it changes the right thing; a confident wrong edit to a shared spreadsheet is worse than a chat reply you can ignore. Beta is the right label until that is proven across messy, real documents.

The second is permissions in practice. "Access matches your Google sharing permissions" is the correct design, but add-ons that read and write live files are exactly the kind of integration that security teams scrutinize, and the controls Anthropic lists will be tested against real compliance requirements rather than a product page.

Our take

As a distribution move, this is sharp. Anthropic is meeting paying users inside the documents they already live in, and it did the unglamorous work of mirroring Google's permission model and carrying its enterprise controls across, which is what makes an add-on like this adoptable at a company rather than just a demo.

As a product claim, treat it as unproven. It is a beta, it is paid-only, and it is walking onto a field where Google's own assistant is already native and free. Use it if you are committed to Claude and want it in your Google files. Do not assume it beats Gemini at the actual editing until someone measures it, because Anthropic has not.