For a while my workflow had a stupid loop in it. Export the CRM list, paste it into Claude, do the work. A week later the list was stale, so I exported it and pasted it again. I was a human sync job, running a manual refresh between two tools that should have just talked. Call it the export-paste treadmill, and I ran on it longer than I want to admit.
Connectors are how you step off it. A connector links Claude to a tool you already use, Slack, Drive, Calendar, your CRM, so it reads the live source and, where you allow it, acts in that system, all from inside a chat.
Under the hood they are MCP, the open standard, packaged as a toggle. That detail matters, and here is why, along with how to turn them on without handing an agent the keys to a live system.
Connectors link Claude to your tools
A connector is a link between Claude and one of your tools, so Claude can read from it and, where allowed, act in it, from inside a normal chat. Instead of exporting data and pasting it in, you let Claude reach the source directly. The tools you live in become things it can see and use, not things you shuttle content out of by hand.
That is the shift, and it is a bigger one than it sounds. It moves Claude from an assistant that talks about your work to one that touches it, which is the whole difference between a demo and a workflow.
They run on MCP
Connectors are the friendly front end for MCP, the open standard covered in its own piece. A connector is an MCP server for a tool, wired once and usable across your Claude surfaces. Knowing that helps, because the same standard powers the connectors in the apps and the tools you add in Claude Code, so what you learn in one place carries to the other.
So a connector is not a bespoke one-off integration. It is the standard plug, turned into a switch you flip. That is why the set of available connectors keeps growing without anyone rebuilding the plumbing each time.
TipIf you understand connectors, you already understand MCP; they are the same idea with a nicer switch. When you outgrow the app, the tools you add in Claude Code work the same way.
How to turn one on
You enable a connector in settings, authorize it to the account you want, and it becomes available in your chats. From there you ask in plain language and Claude uses the connector when the task calls for it. Scope each one to what it actually needs rather than granting everything by reflex.
Start with one, not ten. Pick the tool you paste from most, connect it read-only, and feel the difference before you wire in the rest. A stack of half-configured connectors is worse than one that actually works.
Read first, write on purpose
The same rule from tools and agents governs connectors, and it is the one to internalize. A connector that only reads, your calendar, a Drive folder, a CRM record, is low risk and worth turning on freely. A connector that can write, change a CRM field, send a message, post an event, is where the value and the danger both live.
So grant read first, live with it, and add write access deliberately, to the one connector whose payoff you can name. And put an approval step in front of anything irreversible, because a write connector plus an over-eager agent is one bad instruction away from a very public mistake.
TipTurn on read-only connectors freely; put an approval gate in front of any connector that can write to a live system, and give it the narrowest scope that does the job.
What you can connect
The connectors that matter for GTM are the tools where your real work already lives.
- Slack: post digests and route approvals to where the team already is
- Google Drive and Calendar: your docs, and what is next on your day
- Your CRM: read a deal, draft the update, log the note
- More over time: because it is MCP, anything with a server can become a connector
TipConnect the tool you paste from most first. The export-paste treadmill you are on right now is the exact thing a connector removes.
Where connectors trip people up
The first trap is the one I lived: not connecting the tool at all and staying on the export-paste treadmill because pasting feels safe and familiar. It is not safer, it is just staler; every paste is a snapshot going out of date the moment you take it.
The second is the opposite, wiring a write-enabled connector to a live system with no approval gate and an eager agent, and then being surprised when something changes that you did not intend. Between those two is the sweet spot: read freely, write deliberately, gate the irreversible.
TipA connector that only reads is almost impossible to regret. The regret lives entirely in unguarded write access, so that is the only place to slow down.
What connectors open up for GTM
This is how Claude does real go-to-market work instead of demos. Connect the CRM and it reads a deal and drafts the update. Connect the calendar and it preps you for what is next. Connect Slack and it posts a digest or routes an approval to where the team already is. Chain a few and you have the automations on this site: read the account, act on it, report back, over connectors.
The mental model is simple: files and vision are for artifacts you hand over once, and connectors are for systems Claude should reach on its own, repeatedly. Standing access, not a one-time upload. That is the line that tells you which one you need.
So start with one read-only connector to the tool you paste from most, and step off the treadmill. Which tool are you tired of exporting?
TipStart with one read-only connector to your most-pasted tool, then add write access to the single step whose payoff you can name. Grow the stack one earned connector at a time.
How to set it up
Open Settings and add the connector
In the Claude apps, go to Settings, then Connectors, and add the tool you want, Slack, Google Drive, Calendar, or your CRM. Each one opens its own authorization screen.
Authorize with read access first
Sign in and grant access. Start with read-only scope where the tool offers it, so the first thing Claude can do is look, not change. You are wiring it to your real data, so let it earn write access.
TipConnect it to your own account with your normal permissions. A connector cannot see anything you could not see yourself.
Test it with a read prompt
Prove it works with a question that only your real data can answer:
"From HubSpot, list my 10 oldest open opportunities with no activity in the last 14 days, with the owner and last-touch date."
Allow writes on purpose, one at a time
Once reads are trustworthy, let it act, but deliberately. Ask for the specific change and approve it, rather than handing it blanket write access to your CRM on day one:
"Add a task to the top 3 of those to follow up this week, and log a note on each that we re-engaged."
Frequently asked questions
What are Claude connectors?
Links between Claude and the tools you use, Slack, Drive, Calendar, your CRM, so Claude reads your real data and takes actions in those systems from inside a chat, instead of you exporting and pasting.
What can I connect Claude to?
Common connectors include Slack, Google Drive, and Google Calendar, plus CRMs and more, with the set growing. Because they run on MCP, anything with an MCP server can become a connector.
Are connectors the same as MCP?
Connectors are the app-friendly front end for MCP. Each is an MCP server for a tool, wired once and usable across Claude surfaces, presented as a toggle instead of a config file.
Are connectors secure?
You authorize each one and scope what it can touch, and read-versus-write governs the real risk. Grant read-only connectors freely, add write access deliberately, and gate anything that changes a live system. Review Anthropic's current data terms for specifics.
How do I turn on a connector?
Enable it in settings, authorize the account, and it becomes available in your chats. Start with one read-only connector to the tool you paste from most, then grow the stack one earned connector at a time.
What is the difference between a connector and file upload?
Files and vision are for an artifact you hand over once. A connector is standing access to a system Claude should read on its own, repeatedly. If you are re-uploading the same export every week, you want a connector.
Can Claude change my CRM through a connector?
Only if you grant a write-enabled connector, and you should gate that behind an approval step. Read connectors are low risk; a write connector plus an eager agent with no gate is where mistakes happen.
How do GTM teams use connectors?
To let Claude do real work on the stack: read a CRM deal and draft the update, prep from the calendar, post a digest or route an approval in Slack. Chained together, they power the site's automations, read the account, act, report back.
Sources & further reading
Claude ships fast. This page was last reviewed Aug 22, 2026; verify time-sensitive details against the official docs above before relying on them.