Claude Connectors: How to Plug Claude Into the Tools Your Team Already Uses (And What MCP Actually Is)
TL;DR
Most teams use Claude as a very smart stranger. It knows a great deal about the world and nothing whatsoever about your world — so every conversation begins with you copying something in from somewhere else. Your CRM. Your inbox. Your accounting system. Your Drive. Connectors end that. A connector is a secure link between Claude and a tool your business already pays for, and once it's on, Claude can read from that tool — and, where you allow it, do things in it — without you pasting anything. There are now hundreds of them in Anthropic's official Connectors Directory, spanning roughly a dozen categories: sales and marketing, finance, productivity, communication, design, data, healthcare, education, and more. Underneath them all sits a single standard called MCP (the Model Context Protocol) — think of it as the USB-C port of AI. Before MCP, every tool needed a bespoke integration. Now anyone can build one connector that works everywhere. The practical consequence for a non-technical team is large and specific: the gap between “AI that talks about work” and “AI that does work” is almost entirely a question of what you've connected. Here's what a connector actually is, what MCP means in plain language, how to switch one on, which one to connect first, and the two you should be careful with.
The Copy-Paste Tax Nobody Budgets For
Watch how your team actually uses Claude and you'll notice most of the effort isn't thinking. It's fetching.
Someone wants Claude to draft a follow-up to a client. First they open the CRM, find the account, skim the last three notes, copy them. Then they open their inbox, find the last email thread, copy that. Then they find the proposal in Drive, download it, upload it into the chat. Then — finally — they type the actual request. Four minutes of fetching for thirty seconds of asking.
Multiply that by everyone on your team, several times a day, and you have a real and completely invisible cost. Worse, the fetching quietly limits what people bother to ask. Nobody is going to hand-assemble six months of support tickets into a chat window just to test a hunch about why churn went up. So they don't ask. The question never gets asked, and the answer never gets found.
That is the problem connectors solve. Not “Claude gets smarter” — Claude is already smart. The change is that Claude gets informed. It can reach the CRM, the inbox, and the file itself, which means the four minutes of fetching disappear, and the questions people used to swallow become questions they can just ask.
What a Connector Actually Is
A connector is a door between Claude and one of your business tools, and you decide whether it's open, and how wide.
Concretely: a connector is a secure, authenticated link between Claude and an outside service — HubSpot, Outlook, Google Drive, QuickBooks, Slack, Salesforce, Notion, Airtable, Jira, Box, and hundreds more. You connect it once by logging into that service through Claude, exactly the way you'd authorise any app. From then on, Claude can reach that tool when you ask it to.
Two properties of that door matter more than anything else, and they're the ones to explain to a nervous team.
Claude inherits your permissions — it doesn't escape them. If you can't see a colleague's private folder in Drive, neither can Claude when it's working on your behalf. A connector doesn't hand Claude a master key to the whole system; it hands Claude your key. Whatever your access already is, that's the boundary. This single fact defuses most of the anxiety in the room when connectors come up.
Reading and doing are separate. Some connectors only let Claude look at things. Others also let it act — send the email, update the record, create the file. That distinction is a deliberate control, not an accident, and on a Team or Enterprise plan an administrator can enforce it centrally: connect the tool, but restrict Claude to read-only. We've written before about what happens when a connector crosses that line — when Microsoft 365 gained the ability to actually send your mail and book your meetings, rather than merely read them. The jump from reading to doing is the biggest single step change in what an AI assistant is, and connectors are where you govern it.
MCP: The Boring Standard That Made All of This Possible
MCP is the reason there are hundreds of connectors instead of a dozen. It's the USB-C port of AI, and you never have to think about it again after this section.
You will keep seeing three letters — MCP, the Model Context Protocol — in AI announcements, vendor emails, and probably a slide your IT team shows you. Here's all a non-technical person needs to know.
Remember the era before USB. Every device had its own weird plug. Printers had one cable, cameras had another, phones had a third, and every manufacturer invented their own. Connecting anything to anything was a small research project. Then a common standard arrived, and the problem simply evaporated: one port, one cable, everything plugs into everything.
AI had exactly the same mess. If a software company wanted its product to work with an AI assistant, it had to build a custom integration — and then build a different one for the next assistant, and the next. Slow, expensive, and the reason your favourite niche tool never had an AI integration.
MCP is the standard that ended that. Anthropic published it openly, and it caught on across the industry rather than staying an Anthropic-only thing. A software company now builds one MCP connector and it works. That's why the directory went from a handful of integrations to hundreds in a remarkably short time — and why the tool your team actually uses, not just the famous ones, increasingly has a connector waiting for it.
The business translation: the plug is standard now, so the constraint is no longer technical. It's whether anyone at your company has bothered to plug anything in.
What's Actually in the Directory
Anthropic maintains an official Connectors Directory — a vetted catalogue, not a free-for-all.
Inside Claude you'll find a browsable directory organised by use case, with categories covering sales and marketing, productivity, data, financial services, communication, design, code, education, healthcare and life sciences, and nonprofit. The names in there are the ones your team already lives in: the CRM, the helpdesk, the file store, the project tracker, the accounting package, the analytics tool.
Two details are worth knowing. First, the connectors listed there are vetted by Anthropic for security and reliability — the directory is a curated shelf, not the open internet. Second, connectors are available across every plan, from Free upward, and they work not just in the Claude chat app but in Claude Desktop, Claude Code, and via the API. This isn't an enterprise-only luxury feature. A one-person consultancy on a Pro plan can connect their calendar and their Drive this afternoon.
And if the tool you need isn't in the directory — because it's your own internal system, or something obscure — a custom connector can be built for it. That's a job for someone technical, but it's a small job now rather than a project, precisely because of the standard.
What This Looks Like on an Ordinary Tuesday
Abstractions don't persuade anyone. Here is the same working day, before and after.
The Monday pipeline review. Before: someone exports the CRM to a spreadsheet, cleans it, pastes the summary into Claude, asks what's at risk. After: “Look at every deal in our pipeline that hasn't moved stage in three weeks, check the last email in each thread, and tell me which three are quietly dying and why.” Claude reads the CRM and the inbox itself. The whole preparation step ceases to exist.
The client email you've been putting off. Before: dig through six months of correspondence to remember what was agreed. After: “Read my email history with this client, find what we promised in the March call, and draft a status update that's honest about the two things we're late on.”
The month-end question. Before: someone in finance pulls a report, someone else interprets it, and by the time it circulates it's stale. After: connect the accounting tool and simply ask, “Which costs grew more than 15% versus last month, and is there an obvious reason in the transaction detail?”
The support pattern nobody had time to find. Before: nobody reads 400 tickets. After: “Go through last quarter's support tickets and give me the five complaints that came up most, with a real quote for each.” This is the category of question that only exists once a connector exists — the work was never hard, it was just too tedious to be worth doing.
Notice what all four have in common. None of them require a new skill, a technical hire, or a change to how anyone works. They require one thing: a door that's open instead of shut.
The Two Places to Be Careful
Connectors are the most useful thing your team isn't using. They also deserve two specific pieces of caution — and no more than two.
Be deliberate about write access. Reading is low-stakes; a wrong summary wastes a minute. Doing is different — a sent email cannot be unsent, and an overwritten record is gone. The sensible pattern for a team that's new to this is to start every connector read-only, live with it for a fortnight, and only then turn on writing for the specific tools where the time saved is obviously worth it. Where you do enable it, prefer “Claude drafts, a human sends.” That single rule preserves nearly all of the speed and almost none of the risk.
Know where your data goes. When Claude uses a connector, the data flows to the connected service and is handled under that service's terms — possibly in a different country. For most business tools you already use daily this is a non-issue, because you already accepted those terms when you bought the tool. But it is worth one conversation with whoever owns data protection at your company before you connect anything holding personal or regulated data. Ten minutes of clarity now prevents an awkward conversation later.
What is not worth worrying about, and is where most of the fear actually lands: connectors don't give Claude some sweeping new level of access to your company. They give it your access, to one tool, that you chose, and that you can disconnect with a click.
What Your Team Should Do This Week
Three steps. The first one takes about five minutes.
1. Connect exactly one tool — the one holding the answers you're always hunting for
Open Claude, click the + button in the chat box (or type /), find Connectors, and browse the directory. Pick the single tool your team goes digging in most — usually the CRM, the file store, or the inbox. Connect it, log in, done. Resist the urge to connect eight things at once; one connector that people actually use beats eight that sit there unopened.
2. Ask it the question you've been too lazy to answer manually
The proof isn't in a demo, it's in a real question — specifically one your team has been avoiding because assembling the data was too tedious. Which clients have gone quiet? What are people actually complaining about? Which project is drifting? That's the moment a connector stops being an IT topic and becomes obviously worth having.
3. Write down your team's read/write rule before you need it
One sentence, agreed once, saves a lot of argument: “Claude may read from any connected tool. Claude may only write in tools we've explicitly approved, and anything it sends outside the company gets a human look first.” Decide it now, calmly, rather than in a hurry after someone's AI-drafted email surprises a client.
FAQ
What is a Claude connector in one sentence?
It's a secure link that lets Claude reach into a business tool you already use — your CRM, inbox, file store, accounting system — so it can read from it, and where you allow it, take actions in it, without you copying and pasting anything.
What does MCP mean, in plain language?
MCP (Model Context Protocol) is the common standard that connectors are built on — the USB-C port of AI. Before it, every tool needed a custom integration with every AI assistant. Now a software company builds one connector and it works. It's the reason there are hundreds of connectors rather than a handful, and it's the only thing about MCP a non-technical person needs to know.
Does connecting a tool give Claude access to everything in it?
No. Claude inherits your permissions in the connected service. If you can't see something in the tool, Claude can't see it either when working for you. It gets your key, not a master key.
Can Claude change things, or only read them?
It depends on the connector and on what you allow. Reading and acting are separate permissions. On Team and Enterprise plans, an administrator can enable a connector organisation-wide while restricting Claude to read-only. Our advice for teams starting out: begin read-only, and where you do enable writing, keep a human between the draft and the send.
Do I need a developer to set one up?
Not for the connectors in the official directory — that's a few clicks and a login, no different from authorising any other app. You'd only need technical help to build a custom connector to an internal system that isn't in the directory, and thanks to the standard that's now a small job rather than a project.
Is this only for big companies on expensive plans?
No. Connectors work across all plans, including Free and Pro, and in Claude on the web, Claude Desktop, and Claude Code. Administrator controls — connecting tools centrally, enforcing read-only, restricting which services can be connected at all — are the part reserved for Team and Enterprise.
Where does my data actually go?
To the service you connected, handled under that service's own terms, which may mean outside your country. For tools your company already uses daily, you've already accepted those terms. Still, have one short conversation with whoever owns data protection before you connect anything holding personal or regulated data.
What should we connect first?
Whichever tool your team spends the most time digging around inside. For most businesses that's the CRM, the shared file store, or email. Connect one, use it for a real question this week, and let the second connector be a decision you make because the first one earned it.
Want help choosing which tools to connect, setting the read/write rules that keep it safe, and getting your whole team past copy-paste and into work Claude can actually do? The Deployed Kickstart gets everyone hands-on in a single day, mapped to the tools you already run. The Partner program keeps expanding what's connected — and what your team can do with it — over time.