# MCP and Connectors Explained for Marketers

> What Claude connectors, MCP and plugins mean for marketers: which tools to connect, permission settings, safe workflows and copyable prompts.

Source: https://mentionx.ai/learn/claude/claude-mcp-connectors-marketers  
Author: Dipuck Jones, founder of MentionX  
Updated: 30 October 2026  
Level: Practitioner

**Connectors let Claude read from and act in the tools your marketing team already uses, such as Google Drive, Gmail, Slack, your CRM or your project tracker. They're built on MCP, an open standard Anthropic created, and they're available on every plan, with Free limited to one custom connector. Plugins on paid plans bundle connectors with Skills into a ready-made setup. Start with read-only access, set write actions to "Needs approval", and connect only what a workflow actually needs.** The payoff is less copying and pasting and more work done with real context.

> **Key takeaways**
>
> - Connectors let Claude read and act in your tools, with your own permissions.
> - MCP is the open standard behind them; custom connectors work on every plan (Free gets one).
> - Set read to Always allow, writes to Needs approval, and deletes to Blocked.
> - Plugins bundle Skills and connectors so a team works the same way.

*Checked 3 November 2026 (CET) against Anthropic help and docs pages. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.*

## What is a connector, in plain terms?

A connector is a link between Claude and another service. Anthropic explains that connectors let Claude "access your apps and services, retrieve your data, and take actions within connected services", and that Claude inherits each person's permissions from the connected service ([Anthropic](https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities)). If you can't see a file in Google Drive, Claude can't see it through the connector either.

Once connected, Claude can bring an app into a conversation on its own when it fits your request, so you don't have to name it each time. You'll find available connectors in the Connectors Directory, where each one lists its uses and whether it can read, write or both.

Web connectors are available to all users on Claude, Cowork, the desktop app and the mobile apps. Custom connectors, which point to any remote MCP server, are available on every plan; Anthropic says Free users are limited to one ([Anthropic](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp)).

## What is MCP, and do marketers need to care?

MCP, the Model Context Protocol, is "an open standard, created by Anthropic, for AI applications to connect to tools and data" ([Anthropic](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp)). You can think of it as a common plug shape. Any tool that offers an MCP server can be connected to Claude, and to other AI apps that support the standard ([MCP](https://modelcontextprotocol.io/introduction)).

For most marketers, MCP matters in two ways. First, it's why the list of tools you can connect keeps growing: many marketing platforms now publish MCP servers. Second, if your company has internal data, such as a product catalogue or a reporting database, your developers can build a remote MCP server so Claude can use it. You don't need to understand the protocol to benefit from it.

*Our view:* the question isn't "which connectors exist?" but "which tasks do I copy and paste for?" Every copy-and-paste habit is a candidate connector.

## Which connectors are most useful for marketing?

It depends on your stack, but these categories come up most often:

- **Files and documents:** Google Drive, OneDrive or SharePoint, so Claude can find briefs, brand guides and past campaigns.
- **Email and calendar:** Gmail or Outlook and calendars, for meeting prep and summarising long threads.
- **Team chat:** Slack or Teams, for summarising channels or drafting updates.
- **Project management:** Asana, Linear, Jira, Notion or similar, to turn plans into tasks.
- **CRM and marketing automation:** to look up accounts, segments or campaign performance.
- **Design:** tools such as Figma or Canva, to reference or create assets.
- **Analytics and data:** warehouses or analytics tools, often via your data team.

Check the Connectors Directory for what's available, and check with your IT team which ones your organisation has approved. On Team and Enterprise plans, an owner usually enables connectors for the organisation first, and each user then signs in to the service.

## How do tool permissions work?

This is the most important setting to understand. Anthropic lets you control each connector's actions under Customize > Connectors > Tool permissions. Permissions are grouped by type, such as read-only tools and write/delete tools, and for each one you choose **Always allow**, **Needs approval** or **Blocked** ([Anthropic](https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities)).

Anthropic gives examples that map neatly onto marketing work: allow Claude to search and summarise email but prevent it from sending messages; allow it to read Google Drive files but prevent it from creating or editing documents. These restrictions only narrow access; they never grant more than the source system permits.

A sensible default for a marketing team:

| Action type | Suggested setting | Why |
| --- | --- | --- |
| Search and read | Always allow | Low risk, high value |
| Create drafts or tasks | Needs approval | Easy to review before it happens |
| Send, post or publish | Needs approval or Blocked | Visible to others and hard to undo |
| Delete | Blocked | Rarely needed, costly if wrong |

Revisit the settings once you trust a workflow, but keep anything customer-facing on approval.

## What can you do with connectors in practice?

Here are workflows that marketers use, with prompts you can adapt.

**Campaign context in seconds.** Instead of pasting a brief, let Claude find it.

```
Find the brief, creative feedback and results deck for our
spring webinar campaign in Google Drive. Summarise the goal,
what we tested, what worked and what we said we'd do
differently. Link each file you used.
```

**Meeting prep.**

```
I'm meeting the events agency tomorrow at 10:00. Check my
calendar invite, our last 3 email threads with them and the
shared event plan in Drive. Give me a one-page brief: open
decisions, outstanding actions (who and when) and 3 questions
I should ask.
```

**Plan to tasks.**

```
Turn the launch plan in this doc into tasks in our project
tool: one task per deliverable, with owner, due date and a
link to the doc. Show me the list before creating anything.
```

The last line matters. Even with "Needs approval" set, asking Claude to show the list first lets you fix owners and dates in one go.

## How do Research and connectors work together?

Research can use your connected tools as well as the web. Anthropic says Research searches the web and connected tools such as Google Workspace, and that during Research Claude can use connector tools automatically without asking for approval each time ([Anthropic](https://support.claude.com/en/articles/11088861-using-research-on-claude)). Research started in the mobile app only searches the web.

That combination is powerful for questions like "What have we learned about selling to healthcare buyers?", where the answer is spread across old decks, emails and the web. It's also a reason to review which connectors are switched on before starting Research.

## What are plugins?

Plugins package everything for a role or workflow. Anthropic says each plugin "bundles skills, connectors, and sub-agents into a single package", so you get a ready-to-go setup instead of configuring each piece ([Anthropic](https://support.claude.com/en/articles/13837440-use-plugins-in-claude)). Plugins are available on all paid plans, and organisations can distribute plugins to their members.

For a marketing team, a plugin might combine your brand-voice Skill, a campaign-brief Skill, and connectors to Drive and your project tool. New team members install one plugin and work the same way as everyone else from day one.

## What are the risks?

Three are worth taking seriously.

**Over-sharing.** Connecting a tool lets Claude reach everything you can reach there. If your Drive includes HR files or customer data, Claude can find them when they're relevant to a request. Use the narrowest account and folders that make sense.

**Untrusted connectors.** Anthropic warns that custom connectors connect Claude to services that haven't been verified by Anthropic, and advises connecting only to servers from trusted organisations ([Anthropic](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp)).

**Instructions hidden in content.** An email, document or web page can contain text that tries to steer an AI assistant. Keep write and send actions on approval, and read what Claude proposes before you approve it, especially when it's working with content from outside your organisation.

## From MentionX data: connected context and AI visibility

**Method.** Read-only aggregate queries over MentionX scans from 22 August to 29 October 2026 (UTC): 2,217 engine runs (2,154 completed) across 18 anonymised projects and six engines.

Connectors also make it easier to bring AI visibility data into your planning, from a tracking sheet or a reporting tool. MentionX's cross-engine aggregates show how engines differ in the prompts we track:

| Engine | Brand mentioned | Answers with no citations |
| --- | --- | --- |
| Gemini | 47.3% | 50.7% |
| OpenAI | 42.9% | 30.2% |
| Perplexity | 31.5% | 4.1% |
| AI Mode | 31.3% | 12.6% |
| Copilot | 26.0% | 47.0% |
| DeepSeek | 24.0% | 99.0% |

*Our view:* the most useful connected workflow for GEO is simple: let Claude read your prompt-tracking sheet and your content plan together, and ask which planned pages address the prompts where you're missing.

**Limits.** A convenience sample of mostly B2B software projects, run through engine APIs, which can behave differently from the consumer apps. Perplexity, Copilot and DeepSeek figures come from a single day (26 October 2026), so treat them as a snapshot, not a trend.

## Worked example: Ledgerline connects its stack

*Ledgerline is a fictional accounts-payable tool. Figures are illustrative.*

Ledgerline's marketing team connects Google Drive, Gmail, Slack and its project tool. Read actions are set to Always allow; creating tasks and drafts is set to Needs approval; sending messages and deleting are blocked.

The first week's biggest win is meeting prep: the events manager stops hunting through threads before agency calls. The second is turning the quarterly plan into tasks, which used to take an afternoon.

When a colleague asks to connect an unofficial MCP server found online, the team checks it with IT first and decides to wait for the vendor's official connector. Ledgerline is fictional and the details are illustrative.

> **Pro move:** Make a 'connector map'. List your team's top ten repetitive tasks, the tools each one touches and whether it needs read or write access. Connect only what appears on the map, and you'll have a clear answer when IT asks why.

## How do you roll connectors out to a team?

Start with one workflow and two or three people, not the whole team and every tool. Agree the permission settings, run the workflow for two weeks, and write down what worked. Then add the next workflow. On Team and Enterprise plans, work with the owner who manages connectors so permissions and approved tools are consistent across the organisation.

## What should you try before the next lesson?

Connect one read-only source you use every day, such as Google Drive, and run the campaign-context prompt above. Then open Tool permissions and set write actions to Needs approval, even if you don't plan to use them yet.

Next: [Claude in Chrome for marketers](https://mentionx.ai/learn/claude/claude-in-chrome-marketers). Or return to the [Claude for Marketers track](https://mentionx.ai/learn/claude/claude-for-marketers).

## How do scheduled tasks use connectors?

On paid plans, scheduled tasks can use the same connectors, Skills and plugins as any Cowork task, and they run remotely even when your computer is asleep ([Anthropic](https://support.claude.com/en/articles/13854387-schedule-recurring-tasks-in-claude-cowork)). That turns a good connected prompt into a routine. A Monday-morning digest of the marketing Slack channels, a weekly summary of campaign results from a Drive folder, or a Friday list of overdue tasks from your project tool are all simple to set up.

Be stricter with permissions for scheduled tasks than for live chats, because nobody is watching while they run. Keep scheduled tasks to reading and summarising where you can, and have them produce a draft or a report for you to review rather than taking actions in other tools. If a task must create something, such as tasks in your project tool, check the first few runs carefully.

## How do you check what Claude did through a connector?

Claude usually shows the tool steps it took in the conversation, such as searching Drive or reading an email. Make a habit of expanding those steps when the answer matters. If Claude summarised "the latest results deck", check it opened the right one; file names like "final_v3" catch out people and AI alike.

The services themselves keep records too. Documents Claude edits show in version history, tasks it creates appear under your name, and messages it drafts sit in your drafts folder. That's helpful for accountability, but it also means anything Claude does through a connector looks like you did it. Treat its actions with the same care you'd give your own.

## What should you avoid connecting?

Anything you wouldn't want summarised into a chat. That might include HR folders, legal correspondence, board materials or customer data that your policy says must stay in specific systems. If your account can reach those, consider using folder-level sharing or a separate account for Claude workflows. On Team and Enterprise plans, ask the owner whether there are organisation rules on which connectors and data are approved, and follow them even when a connector would be convenient.

## How do you add a custom connector, step by step?

If the tool you use isn't in Claude's connector directory but offers a remote MCP server (many marketing and analytics tools now publish one), you can add it yourself. Anthropic's steps for individual users are ([Anthropic](https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp)):

1. Go to **Customize > Connectors** in Claude.
2. Click **+ Add**, then **Add custom connector**.
3. Name the connector and paste the tool's remote MCP server URL, which you'll find in the tool's own documentation.
4. Review the authentication settings Claude detects, then sign in.

On Team and Enterprise plans, owners add custom connectors in Organization settings, and members then connect their own accounts. Free users are limited to one custom connector.

Anthropic is direct about the risk: custom connectors connect Claude to services "that have not been verified by Anthropic", and it advises connecting only to servers built and hosted by organisations you trust. A connector from your analytics vendor's official docs is a different proposition from one found in a forum post.

## Which connector workflows are worth copying?

These prompts assume the named connectors are on for the chat. Adjust to your stack.

**Monday campaign digest** (Drive, Gmail, Slack):

```
Look at last week's activity for the [campaign] campaign:
- Docs in the "[folder]" Drive folder changed since Monday
- Emails with [agency/partner domain]
- Messages in #[channel]
Summarise: decisions made, open questions, deadlines this week,
and anything waiting on me. Link each item to its source.
```

**Meeting prep** (Calendar, Gmail, Drive):

```
I'm meeting [name] from [company] on [date]. Find our last
3 email threads, any shared docs, and notes from previous
meetings. Give me a one-page brief: context, what they asked
for last time, what we promised, and 3 questions to ask.
```

**Asset finder** (Drive):

```
Find the most recent approved versions of our [product] one-
pager, pricing sheet and logo files. For each, give the link,
last edited date and who edited it. Flag duplicates or files
that look like older drafts.
```

**Read-only first.** When you try a connector workflow for the first time, phrase it as a read-and-report task. Once you trust the result, add actions such as "draft a reply" with approval still on.

## What should you do when a connector doesn't work?

| Problem | What to check |
| --- | --- |
| Claude doesn't use the connector | Is it turned on for this chat in the tools menu? Name it in the prompt: "Use the Google Drive connector to..." |
| Research ignores connectors | Research started in the mobile app searches only the web; start it on web or desktop |
| "Not authorised" errors | Reconnect the account; sign-ins expire or are revoked by IT |
| Results miss obvious files | The connected account may not have access; check sharing on the file itself |
| Team member can't see a connector | On Team and Enterprise, an owner may need to enable it for the organisation |
| Wrong or partial data | Ask Claude to list the exact searches or tool calls it made, then narrow the prompt |

The last row is the most useful habit. Claude usually shows the tool steps it took, and asking it to restate them in plain language reveals whether it searched the wrong folder, the wrong date range or the wrong account.

## Common questions

**Are Claude connectors free?** Web connectors are available to all users, and custom connectors work on every plan, with Free limited to one custom connector. Plugins need a paid plan.

**Can I stop Claude sending emails through a connector?** Yes. Under Customize > Connectors > Tool permissions, set send actions to Needs approval or Blocked while allowing read actions.

**What is MCP?** The Model Context Protocol is an open standard created by Anthropic for connecting AI applications to tools and data. Claude's connectors are built on it.

## Sources

- Anthropic Help Center, [Use connectors to extend Claude's capabilities](https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities)
- Anthropic Help Center, [Get started with custom connectors using remote MCP](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp)
- Model Context Protocol, [Introduction](https://modelcontextprotocol.io/introduction)
- Anthropic Help Center, [Use plugins in Claude](https://support.claude.com/en/articles/13837440-use-plugins-in-claude)
- Anthropic Help Center, [Using Research on Claude](https://support.claude.com/en/articles/11088861-using-research-on-claude)
All accessed 3 Nov 2026.

## Frequently asked questions

### Are Claude connectors free?

Web connectors are available to all users, and custom connectors work on every plan, with Free limited to one custom connector. Plugins need a paid plan.

### Can I stop Claude sending emails through a connector?

Yes. Under Customize > Connectors > Tool permissions, set send actions to Needs approval or Blocked while allowing read actions.

### What is MCP?

The Model Context Protocol is an open standard created by Anthropic for connecting AI applications to tools and data. Claude's connectors are built on it.

---

Written and reviewed by Dipuck Jones for MentionX Learn. The HTML version of this page: https://mentionx.ai/learn/claude/claude-mcp-connectors-marketers