Anthropic put Claude inside Google Docs, Sheets and Slides in public beta
Anthropic has put Claude for Google Workspace into public beta on every paid Claude plan, with a sidebar that reads the open Docs, Sheets or Slides file and applies edits in place, plus connectors that let Claude create and edit Google files from the chat side.
Image: Anthropic
Why it mattersA model that reads the current selection and writes edits back as approval cards changes where the content work happens, from a prompt-and-paste loop in a chat tab to the file the team already shares.
A chat window beside a Google Doc is one more alt-tab in a workday. Anthropic's announcement, posted on 6 October 2026 and picked up on Hacker News with 85 points yesterday, moves Claude into the Google file itself: an add-on that opens in a sidebar in Docs, Sheets and Slides, reads what you have selected, and applies edits in place.
Claude for Google Workspace is in public beta on every paid Claude plan. The reverse pairing, called the Google Docs, Sheets and Slides connectors, is also in beta: start in Claude and it creates and edits the Google file from the chat side, with Claude's access matched to the sharing permissions on the file.
What the sidebar actually does
In Docs, Claude can fix a sentence or restyle a heading in place without touching the surrounding formatting, and offer bigger rewrites as suggestion cards in the sidebar, each highlighting the passage it would change so the user can apply or dismiss it. Anthropic's example in the post is tightening the executive summary of a product requirements document and turning the next-steps list into a table, both in one request.
In Sheets, Claude can write formulas, build pivot tables and native Sheets charts, and add new tabs. For a join or a data cleaning pass, it pulls the range into Python and writes the result back into the sheet. Anthropic names a Q3 budget versus actuals report with a tab for each team and a summary chart as a worked example.
In Slides, Claude builds new slides from the deck's own layouts and themes so they match the rest, then checks its own work for overlapping elements, text that runs off a slide, or passages that are hard to read.
Approval cards and the "Accept all" switch
The default edit mode is "Ask before edits": each change arrives as an approval card with a summary, and Claude waits for the user to accept it before writing to the file. "Accept all edits" is a one-click swap in which Claude works through a task and applies changes without stopping. Both modes are set per task, inside the sidebar.
Admin, keys and the enterprise controls
Admins deploy the add-on from the Google Admin console to a domain or to selected groups. On Team and Enterprise plans, an owner or primary owner has to turn on the connectors before anyone can start creating Google files from Claude. Anthropic's post names three Enterprise controls that still apply to the add-on: the Compliance API, customer-managed encryption keys, and OpenTelemetry audit export.
When the connectors come up
Starting from Claude, with the Docs, Sheets and Slides connectors enabled, makes sense when the user needs a new file or the work spans several of them: paste a Google file link or ask for a new document, and on a supported setup it opens in a pane beside the conversation. The same Claude connectors and skills a user already has in the chat follow them into the sidebar, so a saved QBR skill or a Salesforce connector works in the Doc too.
The pattern worth noticing
The sidebar reads a user's selection and writes back into the user's own file is the version of in-document AI that leaves the audit trail inside Google's tools: edits land as diffs, approval cards sit beside the text, and the file's existing version history records what changed and when. For a team that already decides what gets published inside Google Workspace, that is where the Claude session should sit too.
Source
Claude now works with Google Docs, Sheets, and Slides, from Anthropic.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.