TL;DR. ChatGPT connectors link ChatGPT to your own apps — Gmail, Google Drive, Outlook, Calendar, Notion, Teams — so it can read and act on your real data inside a chat. As of June 5, 2026 it can also send email. OpenAI renamed connectors to “apps” in December 2025, but the search term stuck.
For a long time, ChatGPT worked in a bubble. It knew a lot about the world and nothing about you — not your inbox, not the file you saved last Tuesday, not what’s on your calendar Thursday. ChatGPT connectors popped that bubble. They let ChatGPT reach into the apps you already use and work with your actual data. The reason people are suddenly searching for what they are: on June 5, 2026, OpenAI turned on the ability to send email straight from a chat — and sending runs on the connector you set up.
ChatGPT connectors are secure integrations that let ChatGPT read and act on the data in your own accounts — Gmail, Google Drive, Outlook, Google Calendar, Notion, Microsoft Teams, and more — from inside a chat. In plain terms: a connector is a permission slip. You sign in, you approve what ChatGPT can see, and from then on it can pull the right email or file into the conversation instead of making you copy-paste it.
Last reviewed: July 1, 2026. Reviewed quarterly, because connector coverage, plan availability, and the send feature are all moving fast right now.
What ChatGPT connectors actually are
ChatGPT connectors are secure, permission-based links between ChatGPT and the apps you already use for work — email, cloud storage, calendars, notes, and chat. Once you connect an app, ChatGPT can read the data inside it and reference that data in its answers, so it stops guessing and starts working from your real information. OpenAI describes them as the way “ChatGPT can use information in the app as context to help provide responses” (OpenAI Help Center, accessed July 2026). You control which apps are connected and, through each app’s login screen, exactly what ChatGPT is allowed to touch.
There’s a naming wrinkle worth clearing up early. OpenAI originally called this feature “connectors,” then on December 17, 2025 renamed the whole area to “apps,” covering both plain data links (Google Drive) and richer interactive tools that outside developers build. So “ChatGPT connectors” and “ChatGPT apps” are the same thing. This guide uses “connectors” because that’s still what most people type into a search box — but if you open ChatGPT and see “Apps” in the menu, you’re in the right place.
Here’s the two-layer version for anyone who wants the mechanism. Under the hood, most ChatGPT connectors are built on the Model Context Protocol (MCP) — an open standard, originally from Anthropic, that lets any AI assistant talk to any tool through a common interface. OpenAI adopted MCP for ChatGPT apps in September 2025, and by early 2026 the wider community had published over 500 public MCP servers, from databases to Slack to email (OpenAI Developers; Wikipedia: Model Context Protocol, accessed July 2026). You don’t need any of that to use a connector — but it’s why the same “connect your tools” idea now shows up in Claude, Gemini, and Copilot. Connectors aren’t a ChatGPT quirk; they’re the shape AI assistants are taking across the board.
What you can connect (and what each one does)
ChatGPT connectors cover the core apps most people run their work life through. The lineup OpenAI documents includes Google Drive, Gmail, Google Calendar, Outlook, Microsoft Teams, SharePoint, Dropbox, Box, Notion, GitHub, HubSpot, and Linear, with more added regularly (OpenAI Help Center, accessed July 2026). The point of connecting each one is the same: give ChatGPT the context that lives inside it, so the answer is about your world, not the internet in general.
Here’s what the most-used ChatGPT connectors let you do:
| Connector | What it does in a chat | Typical use |
|---|---|---|
| Gmail | Search, read, and summarize your email; draft and (new) send replies | “Summarize the thread with the vendor and draft a reply” |
| Outlook (email + calendar) | Search and reference Outlook mail; check calendar; draft and send | “What did the client agree to in last week’s emails?” |
| Google Drive (with sync) | Reference and summarize your Docs, Sheets, and Slides | “Pull the key numbers out of the Q2 review deck” |
| Google Calendar | Read your schedule, find gaps and conflicts | “When am I free for a 45-minute call next week?” |
| Notion | Read pages and databases you point it at | “Summarize the project brief in Notion” |
| Microsoft Teams / SharePoint | Reference messages and files across your org | “Find the latest version of the pricing doc” |
| Dropbox / Box | Reference documents in cloud storage | “Summarize the contract in this Box folder” |
Two honest caveats belong next to that table. First, availability differs by plan — team connectors (SharePoint, Teams, the deeper Business integrations) light up on ChatGPT Business and Enterprise, not on a personal Plus account. Second, most connectors are read-oriented by design; they pull information in. The one big exception is the new ability to send email — the change driving all the fresh attention, so it gets its own section next.
How the new “send email” feature works
The headline change of June 5, 2026 is that ChatGPT can now send email, not just draft it — and it does so through your connected Gmail or Outlook connector. When you ask ChatGPT to write an email, it no longer dumps text into the chat for you to copy. It opens a clean “writing block” that looks like a real email editor. You edit it in place, and if your Gmail or Outlook connector is set up, a Send button fires it off without you ever opening your inbox (Digital Trends; TechRadar, accessed July 2026).
This is the moment the ChatGPT connector graduates from “read my stuff” to “act on my behalf,” so the limits matter. The send feature is available on Plus, Pro, Business, and Enterprise — not Free or Go. It works only in the ChatGPT web app, only with Gmail and Outlook, and it can’t send attachments yet, which reviewers flagged as a real gap (TechRadar, accessed July 2026). We cover the full send workflow — the setup, six prompts worth saving, and the one rule that keeps you out of trouble — in our companion guide, Write and Send Emails With ChatGPT.
The mental model: with a ChatGPT connector, drafting is free and universal, but sending is the new, gated, act-on-your-behalf part.
Are ChatGPT connectors free? What each plan gets
ChatGPT connectors are not free — they are a paid feature, and the newest capability is gated tighter than the rest. The ChatGPT Free tier does not include connectors at all. Connecting Gmail, Drive, Outlook, and the rest is part of the paid consumer plans (Plus and Pro) and the team plans (Business and Enterprise). The email send feature added June 5, 2026 is narrower still, so it’s worth seeing the two layers before you assume your plan can do what a demo showed.
Here’s how access breaks down across the main plans:
- Free — No connectors. You can still paste text into ChatGPT manually, but it can’t reach into your apps.
- Go — No connectors, and no email send.
- Plus (about $20/mo) — Connectors available. Email send available (web only).
- Pro — Connectors available. Email send available (web only).
- Business — The full connector ecosystem, including team apps like SharePoint and Teams, plus admin controls. Email send available.
- Enterprise — Everything in Business, plus data-residency and higher usage limits.
One detail that matters more than pricing: on Business, Enterprise, and Edu plans, OpenAI states it does not use data accessed through connectors to train its models by default (OpenAI Help Center: Admin Controls, Security, and Compliance, accessed July 2026). On the consumer plans (Free, Plus, Pro), the training picture is less clear-cut, which is one reason the safety section below matters. FindSkill’s plain-language take: if you’re connecting work data, the team plans give you the cleaner default. If you’re a solo user on Plus, you get the same features but should be more deliberate about what you connect.
What this means for customer support teams
For a support rep, ChatGPT connectors change the most tedious part of the job: hunting for context before you can even start a reply. A single ticket often means opening the customer’s email thread, checking the order in another tab, and skimming the last three conversations to remember what was promised. With a Gmail or Outlook connector, ChatGPT can read the thread, summarize the history, and draft a grounded reply in one move — and since June 5, send it, once you’ve read it.
The concrete workflow: a support agent on ChatGPT Plus connects Gmail, then types, “Summarize my email history with this customer, then draft a warm reply that acknowledges the delay and offers the two options we agreed on.” The connector reads the actual thread; the draft references real details, not invented ones. Our companion piece on connecting ChatGPT to Gmail for support walks through this setup for a solo rep.
The honest limit: the connector can only summarize what’s in the email. It doesn’t know your refund policy, your tone rules, or the promise you made on a phone call that never got written down. A grounded draft is a starting point you still shape — it doesn’t replace judgment about what the customer actually needs.
The next step: If you want reps who write calm, on-brand replies whether or not they use a connector, our Professional Email Writing course builds that skill in a handful of lessons. Two lessons are free.
What this means for small business owners
For an owner-operator, ChatGPT connectors are closest to hiring a very fast assistant who has already read your inbox. Running a small business means living in your email, your calendar, and a pile of documents — and being the only person who can find anything in them. A ChatGPT connector lets you ask across all of it: “What did I quote this customer last month?”, “Do I have a gap Thursday afternoon?”, “Pull the numbers out of last quarter’s spreadsheet.” No tab-switching, no digging.
The workflow that saves the most time is triage plus drafting. Connect Gmail and Calendar, and ChatGPT can turn “reply to the three emails that need me today” into three grounded drafts, then check whether a proposed meeting actually fits your week. On a paid plan, it can send those replies from the chat once you’ve read them — the write-and-send flow we detail in Write and Send Emails With ChatGPT.
The honest limit for a small business: a connector amplifies whatever’s in your accounts, including the mess. If your inbox is chaos and your files are unlabeled, ChatGPT’s summaries inherit that chaos. And the send feature is powerful precisely because it’s fast — which is exactly why you read before you send.
The next step: To go from “I connected Gmail” to actually running your operation through AI, ChatGPT for Business shows the workflows that save 40 to 60 minutes a day. First two lessons free.
What this means for marketers
For marketers, ChatGPT connectors close the gap between “I have a great idea” and “I have the context to execute it.” Campaign work is scattered across a brief in Google Drive, a thread with the client in Gmail, a content calendar in Notion, and a deck someone shared in Teams. A ChatGPT connector lets you pull all of that into one chat: “Summarize the campaign brief in Drive, check what the client asked for in email, and draft three subject-line options that match.” The AI is finally working from your materials instead of a blank page.
The workflow that lands is context-rich drafting. Point ChatGPT at the real brief through a Drive connector and the copy it produces reflects the actual positioning, audience, and constraints — not a generic guess. Add the Gmail connector, and it can fold in what the client literally said they wanted. The output is a stronger first draft because the inputs are real.
The honest limit: connectors give ChatGPT your context, not your taste. It can match a brief; it can’t tell you the brief is wrong, or that the “safe” subject line is the boring one. The judgment about what’s good — and the instinct to write the line the client didn’t ask for but will love — stays yours.
The next step: The difference between a mediocre and an excellent connector-fed draft is almost always the prompt. Our Prompt Engineering course teaches the patterns that make AI output usable on the first try, across ChatGPT, Claude, and Gemini.
What this means for accountants and finance
For accountants, ChatGPT connectors are useful in a careful, bounded way — and the boundaries matter more here than anywhere. A Google Drive or Outlook connector can pull figures out of a review deck, summarize a long client email thread, or find the right version of a working paper without you searching. That’s real time saved on the coordination and reading that surrounds finance work, especially around month-end and client season.
The workflow that fits the profession is summarize-and-locate, not calculate-and-send. Connect Drive, and ask ChatGPT to “summarize the variance commentary in the Q2 pack” or “find the email where the client confirmed the accrual.” It reads your documents and points you to the answer. What you don’t do is let it invent numbers or fire off a figure-laden client email without checking every line — a fluent, confident wrong number is worse than no number.
The honest limit — and it’s a hard one in finance: connectors read what’s in the file, but ChatGPT can still misread, misattribute, or hallucinate a detail. For regulated, client-facing, or audit-relevant work, a connector is a research assistant you verify, never a signer. Connect the minimum, prefer read-only, and treat every output as a draft.
The next step: Before you point AI at any client data, it pays to understand what these tools can and can’t be trusted with. AI Fundamentals gives non-technical professionals that grounding in a short, plain-language course. Two lessons free.
Common misconceptions about ChatGPT connectors
ChatGPT connectors are new enough that a lot of confident-sounding claims about them are half-right at best. Here are the ones worth correcting — the misunderstandings are exactly where people either miss the value or walk into a risk.
“Connectors mean ChatGPT can read my whole account whenever it wants.”
Not quite. A ChatGPT connector only works after you connect the app and approve permissions, and it can only access what your own account can — OpenAI states connectors “respect your organization’s existing permissions” (OpenAI Help Center, accessed July 2026). The fair worry is the reverse: because it inherits your access, a connector can surface sensitive data you can technically reach but shouldn’t casually point an AI at. The fix is scoping, not fear.
“If ChatGPT can send email now, it can just email people for me automatically.”
Half-true, and the false half is the important one. The June 5, 2026 send feature drafts in a writing block and shows you a Send button — you still review and press it. It doesn’t autonomously email your contacts, it can’t attach files, and it only works on Gmail or Outlook, on the web, on a paid plan (TechRadar, accessed July 2026). A ChatGPT connector accelerates the send; it doesn’t remove you from the loop.
“Connectors and apps are two different things I have to choose between.”
No — they’re the same feature, renamed. OpenAI relabeled “connectors” as “apps” on December 17, 2025 to cover both data-only links and interactive third-party tools (OpenAI Help Center, accessed July 2026). If a tutorial says “connectors” and your menu says “Apps,” you’re looking at the same thing. Don’t burn time hunting for a separate feature.
“Connectors are a ChatGPT thing.”
Only in branding. The underlying idea — an AI assistant that plugs into your tools through the Model Context Protocol — now runs across Claude, Gemini, and Microsoft Copilot too. ChatGPT connectors are one instance of an industry-wide shift, which is why learning the pattern once pays off no matter which assistant you land on.
Related terms
ChatGPT connectors sit inside a cluster of closely related 2026 concepts — the protocol they run on, the rival assistants doing the same thing, and the multi-step pattern that turns reading into acting. These are the terms worth knowing next.
- MCP (Model Context Protocol) — the open standard that most ChatGPT connectors are built on
- Personal Intelligence — Google Gemini’s version of reading your Gmail, Calendar, and Photos for context
- AI Browser — browsers that read pages and act, a close cousin of connectors that reach into apps
- AI Memory — how ChatGPT remembers you across chats, separate from what a connector reads live
- Copilot Agent — Microsoft’s equivalent, a custom AI wired into your work apps
- Agentic AI — the multi-step pattern that turns a connector from read-only into act-on-my-behalf
See also
Courses, ready-to-use AI skill templates, blog posts, and profession hubs connected to ChatGPT connectors — grouped by content type so you can scan for the format that suits you. Pick a course for depth, a skill for a prompt to paste today, and a blog for current context.
Courses on ChatGPT connectors and adjacent topics
- Professional Email Writing — Craft emails that get opened, read, and answered, with or without a connector
- ChatGPT for Business — Save 40 to 60 minutes a day with ChatGPT and Claude
- Prompt Engineering — The prompt patterns that make connector-fed output usable first try
- AI Fundamentals — What these tools can and can’t be trusted with, in plain language
- AI for Gmail — Every AI feature in Gmail, plus the privacy controls that matter
- ChatGPT Workspace Agents for Non-Engineers — Agents that send email, post to Slack, run on a schedule
- AI for Google Workspace — Gemini across Gmail, Docs, Sheets, Slides, and Drive
- MCP Tools Mastery — The protocol connectors run on
- Manage Your Money with ChatGPT — Connect a bank to ChatGPT safely, then budget from real data
- ChatGPT for Excel: 6 FP&A Workflows — Install the sidebar and ship documented finance workflows
- ChatGPT vs Claude — Compare the two head-to-head, two lessons free
- ChatGPT Without the Liability — The privilege-safe AI workflow for HR and SMB owners
Related terms in this glossary
- MCP (Model Context Protocol) — The open standard connectors are built on
- Personal Intelligence — Gemini reading your Gmail, Calendar, and Photos
- AI Browser — Browsers that read pages and act for you
- AI Memory — How AI remembers you between chats
- Copilot Agent — A custom AI you build in Microsoft Copilot Studio
- Agentic AI — Multi-step AI that acts, not just answers
- Context Window — How much an AI can read at once, in tokens
- WebMCP — Letting a website hand AI agents real tools
- AI Hallucination — When a model states something false with confidence
AI Skills (ready-to-use prompt templates)
- Email Response Generator — Craft professional replies in seconds
- Email Tone Adjuster — Shift any email from angry to professional in one pass
- Email Subject Line Optimizer — A/B-test subject lines to lift open rates
- MCP Google Drive Connector — Wire an AI assistant to Google Drive via MCP
- MCP Notion Bridge — Connect an AI assistant to Notion workspaces via MCP
- MCP Slack Connector — Connect an AI assistant to Slack via MCP
- MCP GitHub Integration — Manage repos, issues, and PRs from an AI assistant
- MCP Custom Server Builder — Build your own MCP server in Python or TypeScript
Related blog posts
- Write and Send Emails With ChatGPT: 6 Prompts + Gmail Setup — The June 5 send feature, six prompts, and the one rule
- Connect ChatGPT to Your Gmail to Answer Customer Emails Faster — A solo support rep’s grounded-reply setup
- Is ChatGPT an AI Agent Now? Workspace Agents vs Custom GPTs — What changed as Custom GPTs give way to Workspace Agents
- ChatGPT Workspace Agents for Sales: Weekly Report in 10 Minutes — Wire ChatGPT into Salesforce, Slack, and Gong
- Connect Claude Code to Notion as an External Agent — The OAuth flow for Notion’s External Agents API
- Gmail Live vs Gemini Daily Brief: Which to Turn On First — Google’s two competing Gmail AIs, compared
- Executive Assistants: 5 Gmail AI Prompts to Save 2 Hours — Five Monday-morning Gmail prompts for EAs
- 7 ChatGPT Prompts to Turn Angry Customer Emails Calm — Copy-paste prompts that de-escalate
Profession hubs
- Learn AI for Small Business — The owner-operator playbook for where to start
- Learn AI for Freelancers — Solo-practitioner workflows
- Learn AI for Accountants — The CPA and bookkeeper guide
Degrees and structured programs
- AI Degree in Prompt Engineering — Engineer prompts that work across Claude, ChatGPT, and Gemini
The bottom line
ChatGPT connectors are the moment ChatGPT stopped being a stranger to your work and started reading your actual email, files, and calendar — and, since June 5, 2026, sending email on your say-so. The winners here aren’t the people who connect everything; they’re the people who connect the right things, prefer read-only where they can, and read every draft before it goes out. Connect one app, learn the workflow, and expand from there.
Frequently asked questions
What are ChatGPT connectors? ChatGPT connectors are secure links between ChatGPT and your own apps — Gmail, Google Drive, Outlook, Google Calendar, Notion, Microsoft Teams, Dropbox, and more. Once you connect an app, ChatGPT can read and reference your real data inside a chat: it can pull the right file, summarize an email thread, or check your calendar without you copying anything in. As of June 5, 2026 it can also draft and send email through a connected Gmail or Outlook account. OpenAI renamed connectors to “apps” in December 2025, but most people still search for and say “connectors.”
Are ChatGPT connectors free? Mostly no. The ChatGPT Free tier does not include connectors. Connecting Gmail, Google Drive, Outlook, and other apps is a feature of the paid consumer plans (Plus and Pro) and the team plans (Business and Enterprise). The newer email send feature, added June 5, 2026, is narrower still: it runs only on Plus, Pro, Business, and Enterprise, only in the ChatGPT web app, and only through Gmail or Outlook — not on Free, not on the Go plan, and not in the desktop or mobile apps.
Are ChatGPT connectors safe to use? They are reasonably safe when you use them deliberately, but they are not risk-free. Connectors authenticate through OAuth, encrypt data in transit and at rest, and respect the permissions your account already has — ChatGPT can only see what you can see. For Business, Enterprise, and Edu plans, OpenAI does not use connector data to train its models by default. The real risks are human: connecting an account full of sensitive data you would not want an AI reading, granting a send scope you do not need, or trusting a summary you did not verify. Connect the minimum, prefer read-only where you can, and check what an AI sends before it goes out.
How do I set up a ChatGPT connector? On a paid plan, open ChatGPT and go to Settings, then Apps (or the Apps entry in the sidebar). Pick the app you want — Gmail, Google Drive, Outlook, Calendar — and select Connect. You’ll be sent through the app’s own login and a permissions screen (OAuth), where you approve what ChatGPT is allowed to access. Some apps, like Google Drive with sync, ask you to enable syncing so ChatGPT can reference your files. Once connected, ChatGPT uses the app automatically when it is relevant, or you can name it in your prompt.
What can ChatGPT connectors actually do? Connectors turn ChatGPT from a blank-slate chatbot into one that knows your context. It can search and summarize your Gmail or Outlook inbox, pull answers out of a Google Doc or a slide deck in Drive, check your calendar and find conflicts, reference a Notion page, or read a message in Teams. As of June 5, 2026 it can also send an email you have drafted and edited, straight from the chat, through a connected Gmail or Outlook account. It still cannot send attachments through that flow, and everything runs on your existing app permissions.
What is the difference between ChatGPT connectors and apps? They are the same feature under two names. OpenAI originally called these integrations “connectors,” then renamed the whole area to “apps” on December 17, 2025, to cover both data-only links (like Google Drive) and richer interactive experiences built by third parties. In everyday use, “ChatGPT connectors” and “ChatGPT apps” mean the same thing — connecting ChatGPT to your Gmail, Drive, Outlook, Notion, and so on. The search term “connectors” remains far more common, which is why this page uses it.
Sources
- OpenAI Help Center, “Apps in ChatGPT (formerly connectors),” accessed 2026-07-01. https://help.openai.com/en/articles/11487775-connectors-in-chatgpt
- OpenAI Help Center, “ChatGPT apps with sync,” accessed 2026-07-01. https://help.openai.com/en/articles/10847137-chatgpt-synced-connectors
- OpenAI Help Center, “Connector use cases and prompts,” accessed 2026-07-01. https://help.openai.com/en/articles/12084614-connector-use-cases-and-prompts
- OpenAI Help Center, “Google App for ChatGPT — Data Controls FAQ,” accessed 2026-07-01. https://help.openai.com/en/articles/10408842-google-app-for-chatgpt-data-controls-faq
- OpenAI Help Center, “Outlook Email and Calendar app for ChatGPT,” accessed 2026-07-01. https://help.openai.com/en/articles/12512241-outlook-email-and-calendar-app-for-chatgpt
- OpenAI Help Center, “Admin Controls, Security, and Compliance in apps (Enterprise, Edu, and Business),” accessed 2026-07-01. https://help.openai.com/en/articles/11509118-admin-controls-security-and-compliance-in-apps-enterprise-edu-and-business
- Digital Trends, “You can now send emails directly from ChatGPT on the web,” accessed 2026-07-01. https://www.digitaltrends.com/computing/you-can-now-send-emails-directly-from-chatgpt-on-the-web-without-leaving-your-conversation/
- TechRadar, “I sent an email without opening Gmail thanks to ChatGPT’s new feature — then I found the catch,” accessed 2026-07-01. https://www.techradar.com/ai-platforms-assistants/chatgpt/i-sent-an-email-without-opening-gmail-thanks-to-chatgpts-new-feature-then-i-found-the-catch
- Varonis, “Security Risks in ChatGPT Enterprise Connectors: How to Prepare,” accessed 2026-07-01. https://www.varonis.com/blog/chatgpt-enterprise-connectors
- OpenAI Developers, “Building MCP servers for ChatGPT Apps and API integrations,” accessed 2026-07-01. https://developers.openai.com/api/docs/mcp
- Wikipedia, “Model Context Protocol,” accessed 2026-07-01. https://en.wikipedia.org/wiki/Model_Context_Protocol
