Claude Connectors 101: plug Notion, Gmail, and Linear into Claude in 10 minutes

7 min read  ·  Connectors setup  ·  Last updated 13 May 2026

Part of Learn Claude Code: The Complete Operator's Guide. For the operator's overview of Skills, Connectors, Cowork, and Artifacts, start there.

You can ask Claude to draft an email. Good.

You can ask Claude to read your inbox, find the three follow-ups you owe today, draft replies in your voice, and queue them in Gmail for review. Better.

The first is chat. The second is chat plus Connectors. This post is the plain-English setup guide for the three Connectors that matter most: Notion, Gmail, and Linear.

If you have read our Skills vs MCP post and the kitchen analogy stuck, Connectors are the kitchen. This post shows you how to actually walk in and turn on the lights.

What a Connector actually does

A Connector is the bridge between Claude and a specific tool you already use.

Once you plug in (say) Notion, Claude can read pages from your workspace, search across them, and create new pages, if you grant it write access. From any chat. No copy-paste, no manual context loading.

The same goes for Gmail, Linear, Google Drive, calendar, and a growing list of others.

Connectors are what the developer world calls MCP servers. The plain-English version: live data, live tools, real access.

The first one to set up: Notion (read-only)

Why Notion as your first connector?

Three reasons.

One, if your team already lives in Notion (meeting notes, client folders, project tracker, knowledge base, SOP wiki), Claude can read all of it once connected. Stop pasting context into chat. Ask "what did we agree with Acme Corp last week" and Claude pulls the meeting note. Ask "what is in our SOPs for new-hire onboarding" and Claude reads the whole page.

Two, Notion's permission model is the cleanest of any connector. You pick exactly which pages or databases Claude can see, down to a single doc. Lowest-risk way to learn what a connector actually does.

Three, read-only first. Claude can search your workspace and summarise. It cannot write, delete, or restructure anything. Perfect for the first week.

If you do not use Notion, the same flow works for Google Drive (skip ahead, the setup is identical), Confluence, OneDrive, or Dropbox. Pick whichever already has your team's documents in it. The point is to connect Claude to the place you already store knowledge, not to switch tools.

Setting up these connectors with a real workspace alongside other operators is one of the most-asked-for things at our Claude Code workshops. Half the room arrives never having clicked into Settings.

1. Open Connectors

In Claude.ai or Claude Desktop: Settings → Connectors. You will see a list of available services.

2. Click Notion, then Connect

You will be redirected to Notion's permission page. This is where you choose what Claude can see.

For the first setup, pick a single Notion workspace and a specific page or database to scope access. Do not give Claude your entire workspace on day one. Start narrow, expand later.

3. Approve and return

Notion bounces you back to Claude. The Connector now shows as connected. Total time so far: about 90 seconds.

4. Test it

Start a new chat. Ask Claude: "What is in the Notion page I just connected? Summarise it in three bullets."

Claude reaches into Notion, pulls the page, summarises. You have a working Connector.

The second one: Gmail (read first, draft later)

Gmail is where the productivity gains start to feel real. But it is also where you should be most careful with write access.

1. Connect

Settings > Connectors > Gmail > Connect. Google's permission page opens.

2. Choose read-only first

Google's screen will show you exactly what Claude is asking for. For your first week, pick "read email" and skip "send email".

You will still get to ask Claude to "read my inbox and draft three replies", but the drafts will appear in chat for you to review, not send themselves.

3. Test it on a real prompt

Try: "Read my inbox from this morning. Group the messages by who they are from. Tell me which three need a response today."

If that works smoothly, you can graduate to draft-and-send later. Most of us stay on read + draft-to-review for a while. The marginal time saving from auto-send is not worth the marginal risk.

The third one: Linear (or your team's ticket system)

For software or product teams, Linear changes how planning chats work. Same setup pattern.

Settings > Connectors > Linear > Connect. Approve in Linear. Test with: "What are the open tickets in our 'Voice Agents' project sorted by priority?"

Once that works, you can ask Claude to draft a sprint plan, write release notes from completed tickets, or summarise this week's blockers. All by reading from the live source.

If your team is not on Linear, the same flow applies to most ticket systems with a Connector: Jira, GitHub Issues, Asana, ClickUp.

The permissions that matter

Every Connector asks for a set of scopes. Claude is upfront about what it wants. Most people skim past this. Read it.

Three habits worth building from day one:

1. Start read-only

Always. Even for tools you trust. The first month is for getting comfortable. Read-only lets you see what Claude does without it being able to change anything.

2. Scope to the smallest thing that works

If Claude can read your "Marketing" Notion page, that is better than reading your entire workspace. If Claude can see your "Inbox" Gmail label, that is better than your full mail history.

You can always widen access later. Narrowing it after a leak is harder.

3. Review the audit log

Settings > Connectors > [tool name] shows you the calls Claude has made. Check it weekly for the first month.

This is also the Karol Zieminski "always ask for a change log" habit applied to Connectors. We covered the change-log discipline in Module 6 of Skills 101.

Three workflows that change the next time you open your laptop

Workflow 1: Monday morning inbox triage

With Gmail read + draft-to-review: "Look at my inbox from the weekend. Group by sender. Tell me the three I need to reply to this morning. Draft replies in my usual short, direct tone."

Five minutes of work becomes 30 seconds. Drafts sit in Gmail for you to read, tweak, send.

Workflow 2: Pull a client one-pager from your notes

With Notion read: "In my 'Acme Corp' Notion database, find all meeting notes from the last 30 days. Summarise the open items, the agreed decisions, and what we promised by when. Write it as a one-page client update."

Half an hour of scrolling and reformatting becomes one prompt.

Workflow 3: Sprint summary for the team

With Linear read: "From the tickets closed in our last sprint, write a 200-word summary of what shipped. Group by area. Note any blockers carried over."

The Friday update writes itself.

Bonus workflow: build a client record from your inbox

This is where Notion plus Gmail get powerful together.

With Gmail read + Notion write: "Find every email from Acme Corp in the last 90 days. Pull the key decisions, requests, and promises. Save it as a new page in the 'Client Records' Notion database with sections for Decisions, Open Requests, and Watch-outs."

Two minutes later you have a structured client record built from your real inbox history. The kind of thing operators promise themselves they will "do properly one day" and never get to.

Connectors plus Skills

Connectors give Claude access to your tools. Skills give Claude the workflow for using them well.

On their own, Connectors are powerful but generic. Add a skill that says "when I ask for an inbox triage, always group by sender first, always flag anything from a client as priority, always draft in my short voice" and now the workflow is consistent every Monday.

The pairing is what we cover in the live workshop. It is also what makes the difference between Claude as a tool and Claude as a coworker.

What to do next

Pick one Connector. Do the 90-second setup tonight. Start with read-only.

Run one real prompt against it tomorrow morning. See what changes.

Then come back and read What is a Claude Skill to layer the workflow side on top. That is where the leverage gets compound.