claude connectors, explained: what to set up first.
the questions arriving in my search data are all versions of the same one: how do i connect claude to my actual stuff? here is the plain-language version.
claude connectors let claude work inside your real tools: your files, calendar, email, and trackers. they are built on MCP, the model context protocol, anthropic's open standard for connecting AI to software. the practical effect: claude stops describing work and starts doing it, with you approving what ships.
what changes when claude is connected
unconnected, claude is advice. connected, it is labor: it reads the actual thread before drafting the reply, checks the actual calendar before proposing times, pulls the actual document before writing the summary. the quality jump is not the model getting smarter. it is the model finally seeing your real context.
which connector first
| connector type | what it gives claude | set up when |
|---|---|---|
| files / drive | your documents as context | first, almost always |
| calendar | real availability and commitments | first week |
| threads to triage and draft against | once approval habits exist | |
| crm / tracker | deal and task state | when follow-up goes systematic |
the order matters because context compounds. files first means every later connector inherits a claude that already knows your business. that is also why the flat setup work, files, not chats, pays off before any connector does.
the safety rules that keep you in control
grant read before write. keep an approval gate on anything that sends, posts, or changes state. review the first week of outputs like you would a new assistant's. connectors are permissioned per tool, so you decide exactly how much hands the system gets, and you can take access back at any time.
takeaway
connectors are the difference between an AI that talks about your work and one that does it. files first, calendar second, approvals always.
skip the trial and error
the vault includes the connector-ready workflows.
every skill in the vault is written to run on your real context: follow-up, documents, content, reporting. install, connect, approve.
see the skills vault →more field notes on x · instagram · tiktok
written by elisabeth hitz, certified in anthropic's ai fluency program (framework & foundations, and ai capabilities & limitations), plus claude 101 and claude cowork.
MCP is anthropic's open protocol; see anthropic's documentation at modelcontextprotocol.io for the technical spec. related: claude for business owners