How to Connect Gmail to Grok bot in 2 Minutes (2026)

Connecting Gmail to Grok takes about two minutes, and the first thing to know is that you are not connecting Gmail.

Grok has no standalone Gmail connector. It has a Google Workspace connector, and xAI's own description of it is that Grok can "read and write emails, documents, spreadsheets, and events." One grant, four surfaces, both directions. If you only wanted inbox summaries, you are approving considerably more than that.

That is not a reason to avoid it. It is a reason to read the consent screen rather than clicking through it.

The setup

Connectors went live on 6 May 2026 across Grok Web, iOS and Android.

On the web: sign in, click the + button, select Connectors, then + Add connector. Pick Google Workspace and complete Google's consent flow.

On iOS or Android: go to Settings → Connectors and do the same.

That is the whole thing. Grok then has access until you revoke it, either from Grok's connector settings or from your Google account's third-party access page, which is the one worth bookmarking because it works even if you lose access to Grok.

If you are on Business or Enterprise, there is a step before all of this: a team admin has to provision the connector in the cloud console first. Members cannot add it themselves until that happens. If the option is missing from your Connectors list, that is usually why rather than a bug.

What you actually granted

Worth being specific, because "connect Gmail" and what the connector does are different sizes.

The other connectors available alongside it are SharePoint, Outlook, OneDrive, Notion, GitHub and Linear. SharePoint is the one with an explicit write toggle: read access lets Grok search and synthesise, and xAI notes that "with write permissions enabled, Grok can also create, edit, and update documents, lists, and pages." Google Workspace is described as read and write without that separate switch, so treat it as write-capable from the moment you connect it.

The three things worth knowing

Your inbox is attacker-controlled content

This is the one that matters and it is not specific to Grok.

Anyone can send you an email. That email is content the assistant will read. If it contains text addressed to the assistant rather than to you, and the assistant also holds a connector that can send mail or edit documents, then reading has become a path to doing.

Published research through 2025 and 2026 demonstrated exactly this shape against mainstream assistants, including a zero-click exfiltration through a research agent's mail connector and a separate hijack delivered through a calendar invite. Both were patched. The underlying pattern is not patchable, because following instructions in text is what these models do.

The practical consequence: do not reason about "read access" as if it were passive. Reason about the combination of everything the assistant can read and everything it can do. We wrote the full version of that argument in a security checklist before connecting any AI assistant to company data.

A connected account is not a permission system

Google's consent screen decides whether Grok may act as you. It does not decide how much it may do.

There is no threshold in that grant. No "drafts are fine, sending to more than ten recipients needs a human." No rule that calendar events may be read but not deleted. You get an on switch, and then the assistant's judgement.

For personal use that is usually acceptable. For anything connected to a shared company mailbox it is the gap to close, and closing it means putting something between the assistant and the action that can evaluate a rule at execution time. That is the argument in policy-as-code for AI agents and human-in-the-loop approval.

Revocation is the thing to test first

Before you rely on the connection, practise removing it. Revoke from Grok's settings, then check Google's own third-party access list to confirm it actually went.

Doing this once, while nothing is wrong, is worth more than reading any amount of documentation about it later.

When you need more than a connector

The built-in connectors cover the mainstream SaaS surfaces. They do not cover your billing system, your internal admin API, or the long tail of tools your team actually runs on.

For those, Grok supports Bring Your Own MCP: you point it at a custom MCP server and it discovers the tools that server exposes. That route also gets you the controls the built-in connectors lack, because you decide what the server allows. We go through it properly, including the part where the server has to be reachable over the public internet, in Grok Bring Your Own MCP.

If what you want is a governed path to a specific provider rather than broad account access, the shape looks like this, with no .env file anywhere:

npm install -g swytchcode
swy login
swy init --mode=sandbox

swy get gmail
swy discover "send an email"
swy info gmail.user.send.create1
swy add method gmail.user.send.create1   # only this operation can run
swy auth connect gmail

The difference from a broad grant is that swy add method produces a finite list. An assistant asking for an operation you never added fails at the runtime rather than reaching Google, and the credential sits in ~/.swytchcode/credentials.db outside your project where the model cannot read it. That is credential brokering.

How it compares to other assistants

Short version: the connector lists differ, the permission models are more alike than not, and none of them gives you conditional authorization out of the box. The longer comparison is in Grok vs ChatGPT vs Claude connectors.

FAQ

Does Grok have a Gmail connector?

Not as a standalone connector. Gmail is reached through the Google Workspace connector, which also covers Docs, Sheets and Calendar in the same grant. xAI describes it as read and write across all four, so connecting it for inbox summaries also grants the ability to send mail and edit documents.

How do I connect Gmail to Grok?

On the web, sign in, click the + button, choose Connectors, then + Add connector and select Google Workspace. On iOS and Android, go to Settings → Connectors. On Business and Enterprise plans a team admin must provision the connector in the cloud console before members can add it.

What can Grok do with my Gmail once connected?

Read and write. That includes searching and summarising mail, and also drafting and sending it, plus reading and editing documents, spreadsheets and calendar events under the same grant. There is no threshold or approval step inside the connector, so it is on or off.

Is it safe to connect Gmail to Grok?

For personal use with nothing sensitive, usually fine. The risk worth understanding is that your inbox is content anyone can write to, and an assistant that reads attacker-supplied text while also holding the ability to send mail has a path from reading to acting. If the mailbox is shared or business-critical, put something between the assistant and the destructive operations rather than relying on the grant alone.

How do I disconnect Grok from Gmail?

Remove the connector in Grok's connector settings, then confirm it is gone from your Google account's third-party access list. Checking both matters, and it is worth doing once as a test before you depend on the connection.

What other connectors does Grok support?

As announced on 6 May 2026: SharePoint, Outlook, OneDrive, Google Workspace, Notion, GitHub and Linear, plus Bring Your Own MCP for custom servers. SharePoint has an explicit write-permission toggle; the others are described by capability rather than by separate read and write switches.

Can I connect Grok to an internal API?

Yes, through Bring Your Own MCP rather than a built-in connector. You add a custom connector pointing at your own MCP server URL, and Grok discovers the tools it exposes. The server has to be reachable over the public internet, which is the part that needs thought rather than a tunnel and a shrug.

Setup commands on this page come from the Swytchcode CLI docs. Connector details are from xAI's own announcement and documentation, described here rather than linked.