Can Claude read my emails?

Short answer: only if you set it up. Claude has no view of your mail until you connect a Google account or paste a message into a chat. Below: what the Gmail connector can read and send, which plans have it, how to connect it, how to take it back in both places, what Anthropic says about training — and what to do when all you wanted was help with one thread.

No — not unless you connect it, and connecting is a step you can't take by accident. Out of the box, Claude knows nothing about your inbox. There are two ways mail gets in: you paste the text of a message (or upload a saved copy, such as a PDF) into a chat, or you connect a Google account through Claude's Gmail connector, which sends you to Google to sign in and approve the permissions. No background access exists before that moment.

That's the easy half. The half worth ten minutes of your time is what the connection covers once you grant it, because it is wider than the task most people had in mind. Everything below about Claude comes from Anthropic's own published pages (and Google's, for the revoke steps), read on 20 September 2026, with links at the end of the help section.

What the Gmail connector is, and which plans have it

A connector is Anthropic's word for a link between Claude and an outside account. The one people search for as "Claude Gmail integration" is part of a set of three: in Anthropic's words, "Google Workspace connectors (Gmail, Google Calendar, and Google Drive) are available for all users on Claude and Claude Desktop." Read plainly, all users includes the free plan, and the same page's training note names Claude Free alongside Pro and Max. One caution: Anthropic's separate Claude Docs page for Gmail (last modified 7 August 2026, ten days before the help article) still says "Available on Pro, Max, Team, and Enterprise plans." The two pages disagree. The help-center article is the newer one, and your own Customize → Connectors screen settles it. And despite the "Workspace" in the name, the page asks only that you "authenticate directly with your Google account" — it states no requirement for a company Workspace subscription.

The one plan-level difference is who decides. On a personal plan, you do. On Team and Enterprise plans, "an Owner or Primary Owner must enable these connectors at the organization level before individual users can authenticate." More on work accounts below.

What Claude can read — and whether it can send

Anthropic's help page lists what the Gmail connector does. Reading first:

• Search and read. "Search and read emails using natural language queries." That is the whole mailbox of the account you connected — the page describes no way to narrow the connection to a label, a sender or a single thread. Claude "mirrors your existing permissions," so its reach is your reach.
• Attachments: names, not contents. "Access email metadata, including attachment metadata (not attachment content)." If the thing you need is inside the PDF a client attached, the Gmail connector won't open it.
• Only when asked. "Claude only accesses your data when you explicitly ask a question or request an action requiring this information, and retrieves the minimum information needed." This is a real limit and deserves credit: the grant reaches everything, but the fetching happens per request.
• It shows its sources. Responses include "citations indicating which emails, calendar events, or documents were used as sources, with links to the originals when available."

Then acting. If you have read that Claude can only draft, or cannot write mail at all, you may have read one of Anthropic's own pages: its Claude Docs page for Gmail (last modified 7 August 2026) still lists "Claude cannot create, send, or modify emails" under Limitations. The help-center article, updated 17 August 2026, says otherwise. The help page's own FAQ asks "Can Claude send emails on my behalf?" and answers: "Yes. Claude can send, reply to, and forward emails from Gmail, and asks for your approval by default before each of these actions. On Team and Enterprise plans, owners decide whether members can allow these actions to run without asking each time." The same list includes "Draft emails with proper formatting and context," "List saved drafts in your Gmail account," and "Manage email organization with labels and threads." The connector's public listing on claude.com names further tools in the same family — moving a message to trash, marking spam, creating a filter. The help page's general rule covers them: "By default, each action Claude takes on your behalf requires your explicit approval."

Two practical consequences. First, Google's permission screen will mention sending, and Anthropic addresses that directly: "Claude can send, reply to, and forward emails, but only does so with your explicit approval by default." Second, note the word default. On Team and Enterprise plans the page says owners decide whether members may let these actions run without asking each time; it does not spell out how that choice appears on a personal plan. Anthropic's general connectors article describes per-tool permissions — Always allow, Needs approval, Blocked — in its section for Team and Enterprise owners, and the Google page says individual users "can also disable specific tools from the chat interface." If you only ever want Claude to read, look for those controls on your own Customize → Connectors screen and block sending outright, rather than relying on remembering to say no.

How to connect Claude to Gmail

You can do this on the web or in the desktop app; Anthropic's general connectors article says installing connectors from Claude's iPhone and Android apps is still being tested, so start on the web if the option is missing:

• Open Customize → Connectors (or, from a chat, press +, hover over Connectors, and choose "Manage connectors").
• Press the + button next to Connectors and pick Gmail from the directory.
• Read the connector's description and capabilities, then press Connect.
• Sign in at Google and approve the permissions. Read that screen. It is the only place the full grant is spelled out, and Anthropic's own wording for this step is blunt: "you're granting Claude permission to access and potentially modify data within that service based on your account permissions."
• Back in Claude, Gmail can be switched on or off per conversation: press + in the chat, hover over Connectors, and toggle it. A connection made on the web or desktop is available in the phone app the next time you log in.

Anthropic lists Gmail, Google Calendar and Google Drive as separate connectors, each with its own directory entry and its own on/off toggle. So wanting help with email should not oblige you to connect the other two, and Google's permission screen shows exactly what is being granted.

How to disconnect Claude from Gmail — in both places

Undoing it deserves the same care as doing it, and the clean version has three parts:

• In Claude. Go to Customize → Connectors, find the Gmail connector (Anthropic's page calls it "your Google connector") and click it to open its details, press Disconnect, then press Disconnect again in the confirmation message. (Anthropic documents these steps in its troubleshooting section; they are the same steps whether you are reconnecting or leaving.)
• At Google. This ends the grant at the source, whatever happens on Claude's side. Anthropic's page sends you to myaccount.google.com/connections: "Search for the Claude connection, click into the settings, and choose 'Delete all connections.'" Google's own help article calls it your Google Account's linked apps page (it links to myaccount.google.com/linkedapps): choose "Access to your Google Account," select the app, then See details → Remove access → Confirm. Google's wording on the effect: "If you remove access, the app can’t access your Google Account." The two sources label the buttons differently and Google reshuffles this page often, so go by what your screen shows.
• Then the chats. Disconnecting stops new retrieval. It does not reach back into conversations that already pulled mail in. Anthropic: "Data retrieved while using connectors is stored on Anthropic servers," and "This data is retained with its associated chat, so you can delete any retrieved data by deleting the chat." Google makes the general point about any linked app: after you revoke access, "You may need to contact the developer of the app to request that they delete the data they already have." Here the route is simpler — delete the chats.

What Anthropic says about training on your email

The statement is direct: "We do not train our models on your Gmail, Drive, or Calendar connector data, ensuring your private information remains private." Anthropic's privacy center says the same from the other direction: the data it may use to improve models "does not include raw content from connectors (e.g. Google Drive)."

There is one carve-out, and Anthropic states it on the same page rather than burying it. If you are on a consumer plan (Free, Pro or Max) and you have chosen to let your chats be used for model training, then "any content you copy/paste from your Gmail, Drive, Calendar, or Claude's responses which include specific information from these connectors may be used to improve our models." In plain terms: what the connector fetches is excluded, but the conversation around it follows your ordinary privacy setting — and Claude's reply summarising a client's email is part of the conversation. If your inbox holds client material, check that setting in Claude's privacy settings before you connect, not after. The page also says data is "encrypted at rest and in transit."

Work accounts: someone else may hold the switch

If the mailbox belongs to your firm, two different administrators can each say no.

• Your firm's Claude plan. On Team and Enterprise, an Owner enables connectors for the organization before anyone can sign in to one, can disable the Google connectors entirely under Organization settings → Connectors, and — per the general connectors article — can restrict actions for everyone. Anthropic's own first example: "Allow Claude to search and summarize email, but prevent it from sending messages." Individual users can't override that.
• Your firm's Google Workspace. If Google shows "Access blocked: your institution's admin needs to review Claude for Google Drive," the block is on Google's side. Anthropic's page explains that a Workspace admin has to mark Claude as a trusted app in the Google admin console (Security → Access and data control → API controls → Manage third-party app access).

A block like that is usually a policy decision, not a glitch. If you hit one, the useful move is to ask what the firm's position on client mail and outside assistants is — before finding a personal-account way around it.

Does it cover Outlook and Microsoft 365?

Partly. There is a separate Microsoft 365 connector, and Anthropic says it "is available on all Claude plans: Free, Pro, Max, Team, and Enterprise." It covers Outlook email and calendar, SharePoint, OneDrive and Teams. Three conditions, all from Anthropic's "Connect to Microsoft 365" article (updated 8 September 2026):

• Work accounts only. "Personal Microsoft accounts (such as @outlook.com, @hotmail.com, or @live.com) can't be used." If your personal mail is on Outlook.com, this connector is not for you.
• An admin goes first. "Before anyone in your organization can connect Microsoft 365, an admin needs to set it up," including a one-time consent from the organization's Microsoft administrator.
• Reading is standard; sending is up to the admin. "Write tools are admin-controlled. Claude can always search and analyze your data." Drafting and sending mail are available only where the admin has enabled them. As with Gmail, access is on request: "It doesn't run background searches."

Disconnecting is the same motion — Customize → Connectors → Microsoft 365 → Disconnect. Other routes into a mailbox exist as well (Anthropic's browser extension, third-party automation services, custom connectors), each with its own permissions; this guide covers the built-in ones.

Is it safe to connect Claude to Gmail? Two separate questions

Is the mechanism sound? Yes, by any ordinary standard. You sign in at Google, not on a form of Anthropic's. Google shows you the grant. Anthropic publishes what the connector does, says it fetches only on request, excludes connector data from training, and gives you a documented way out on both sides. A connector is a legitimate, revocable grant, and for plenty of people it is the right tool — if what you want is an assistant that can search ten years of mail for something you never thought to save, nothing narrower will do that.

Is the scope right for what's in your inbox? That one is yours, and no vendor can answer it for you.

• The grant is the mailbox. You wanted help with one negotiation. The permission covers every message you have kept, from every client, plus whatever personal mail shares the account.
• Most of your inbox is other people's information. If you work client-side — consulting, legal, finance, medicine — those threads were sent to you in confidence, often under an engagement letter or an NDA with its own terms about third-party services. None of those people were asked.
• It can act, not only read. Approval-by-default is a good default. It is still a default, and on a work plan the owner decides whether it can be switched off.
• What was fetched stays with the chat until you delete the chat.

A workable rule: connect the account only if you would be at ease with every message in it being reachable, and the terms you work under permit it. If you hesitated at "every message," send the one thread instead.

Sources, all read 2026-09-20. Anthropic help center: Use Google Workspace connectors (updated 2026-08-17), Use connectors to extend Claude's capabilities (updated 2026-08-20), Connect to Microsoft 365 (updated 2026-09-08). Claude Docs: Gmail integration (last modified 2026-08-07; conflicts with the help center on plans and sending, as noted above). Anthropic privacy center: Is my data used for model training? (updated 2026-03-16). Connector listing: claude.com/connectors/gmail. Google Account Help: Manage links between your Google Account & apps from other developers and Share some access to your Google Account data with apps from other developers. Vendors rename and move these screens often; where your own settings disagree with this page, your settings are right.

The question underneath: you wanted help with one thread

Very few people typing "can Claude read my emails" want an assistant with standing reach into all of them. The real request is smaller and entirely reasonable: there is one thread — the pricing back-and-forth with a client, the contract redlines, Thursday's agenda — and I'd like my assistant to have read it before I walk into the call.

A connector answers a different question. On scope, its two settings are not connected and the mailbox. Pasting a message into a chat is the narrow alternative, and it works for a one-off — but a live thread keeps moving, and the copy you pasted on Tuesday is stale by the time the client's CFO replies on Thursday evening.

The narrow request needs something else: an assistant you can simply send the thread to. Anthropic's help pages describe no address you can forward mail to; the routes they describe are connecting an account or pasting the message in.

CC your assistant. It reads the thread — not your inbox.

Equerry is a personal assistant that works the way you would work with a human one: you copy them on what matters. Your Equerry has its own email address — something like mark@mail.getequerry.app, added to your contacts when you set it up — and what it knows about your email is what you sent to that address. There is no mailbox sign-in, nothing to approve at Google, and nothing to revoke later, because it has no way into your inbox in the first place. The send button is the whole arrangement.

• Forward an email. Ask about it later. The scope document a client sent on Tuesday, the engagement letter, the revised timeline — forward it once and it can be asked about next week, or next quarter.
• CC it on the threads that matter. Add your Equerry address to a live client thread and, for as long as it stays on the CC line, each later reply arrives on its own — including the one that lands at 6 PM while you are in another client's session.
• Walk in current. Ask "where did we leave pricing with Acme?" before the 9 AM call and the answer reflects Thursday's reply, not Tuesday's paste.
• The original stays where it was. Your email remains in your own mailbox, untouched. Equerry works from the copy you sent it and answers from what that thread said, so the message itself is always there to check against.

Three honest notes. The narrow path has a cost: Equerry can't go looking for the email you forgot to send it, so a mailbox connector and a CC'd assistant are good at different jobs — plenty of people will keep Claude for drafting and thinking, and use this for the client record. A CC is visible: everyone else on the thread can see your Equerry address on the CC line, so expect the occasional question about it. And forwarding a thread is still sharing it: what you send is processed to make it searchable and answerable, including by the outside providers named on our AI disclosure page. The professional question doesn't vanish. It shrinks to a size you can actually answer — one thread you chose, not the archive.

If email is where your clients live, the next reads are how to keep track of important emails and how to find an old email you can't remember. If you're weighing the same decision for the other assistant on your phone, see can ChatGPT read my emails?

Last updated 2026-09-21.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.