Claude Cowork Plugins for Marketers: The Complete Implementation Guide

Cowork plugins turn Claude from a generic AI assistant into a marketing operations specialist. Released January 30, 2026, plugins bundle skills, connectors, slash commands, and sub-agents into packages tailored to specific job functions. Marketing teams now connect Claude directly to HubSpot, Semrush, Google Analytics, monday.com, and Microsoft Clarity through Model Context Protocol (MCP) servers.

The marketing plugin alone ships with connectors for Slack, Canva, Figma, HubSpot, Amplitude, Notion, Ahrefs, SimilarWeb, and Klaviyo. This guide shows you how to install, customize, and build automation workflows using these tools.

TL;DR

Cowork plugins connect Claude to your marketing stack through MCP servers. Install the marketing plugin from Claude Desktop in one click. Customize it by adding your brand voice, campaign processes, and tool credentials. Build workflows combining HubSpot CRM data, Semrush keyword insights, GA4 analytics, and monday.com task management. Enterprise teams report 15-20 hours saved weekly on reporting and campaign setup tasks.

Key Takeaways

Cowork plugins bundle skills, MCP connectors, and slash commands into role-specific packages for marketing teams
The marketing plugin ships with pre-built connections to 9 major platforms including HubSpot, Amplitude, and Ahrefs
Install plugins from Claude Desktop’s plugin marketplace or upload custom plugins built with the Plugin Create tool
MCP servers like Semrush, Microsoft Clarity, and monday.com require separate authentication but work inside Cowork
Customization happens through plain markdown files. No code required. Edit skills to match your brand voice and processes
Plugins run locally on your machine with organization-wide sharing coming in future updates
Enterprise workflows combine data plugin SQL queries with marketing plugin campaign analysis for cross-functional insights

What Are Cowork Plugins and How Do They Work?

Cowork plugins are file-based packages containing four components: skills, connectors, slash commands, and sub-agents. Skills encode domain expertise Claude draws on automatically. Connectors wire Claude to external tools via MCP servers. Slash commands trigger explicit actions like /marketing:campaign-brief. Sub-agents handle specialized tasks within larger workflows.

Every plugin follows this structure:

marketing/
├── .claude-plugin/plugin.json   # Manifest file
├── .mcp.json                    # Tool connections
├── commands/                    # Slash commands
└── skills/                      # Domain knowledge

The file-based architecture means customization requires no code. Edit markdown files to change how Claude handles your workflows. Add your terminology, org structure, and campaign processes directly into skill files.

Action item: Open Claude Desktop, navigate to the Cowork tab, click Plugins in the left sidebar, and install the Marketing plugin. Takes 30 seconds.

What MCP Connectors Does the Marketing Plugin Include?

The marketing plugin ships with 9 pre-configured connectors targeting content, analytics, and campaign management platforms. Each connector uses MCP to give Claude read and write access to your tools.

Default marketing plugin connectors include:

Slack for team communication and campaign notifications
Canva for design asset management
Figma for design collaboration
HubSpot for CRM and marketing automation
Amplitude for product analytics
Notion for documentation and knowledge bases
Ahrefs for SEO and backlink analysis
SimilarWeb for competitive traffic intelligence
Klaviyo for email marketing automation

These connectors activate after you authenticate each service. Claude respects your existing permissions. If you have read-only access to HubSpot deals, Claude has read-only access to HubSpot deals.

Action item: Check your team’s API access for each platform. Document which scopes each team member needs for their specific workflows before rolling out Cowork.

How Do You Connect Semrush to Claude Cowork?

Semrush provides an official MCP server at https://mcp.semrush.com/v1/mcp. This server gives Claude access to keyword research, competitor analysis, and backlink data through the same API your browser accesses.

Add the Semrush MCP connector in Claude:

Navigate to Settings → Connectors → Add custom connector
Name the connector “Semrush”
Enter the server URL: https://mcp.semrush.com/v1/mcp
Save and complete OAuth authentication when redirected

Semrush MCP provides data from Domain Analytics, Keyword Analytics, Backlink Analytics, and the Trends API. Your access depends on your Semrush subscription level. Standard API units apply to all requests.

Sample prompts after connecting Semrush:

SYSTEM: You are a marketing analyst with Semrush access.

Pull the top 20 organic keywords for {{COMPETITOR_DOMAIN}}.

Compare keyword difficulty and search volume.

Output: Table format with recommendations for which keywords to target.

The Semrush MCP server supports retrieving a domain’s 12-month advertising history including keywords bid on, ad positions, and ad copy. Use this for competitive PPC analysis without manual exports.

Action item: Connect Semrush MCP and run your first query: “Show me the top organic keywords for [your main competitor]”. Compare against your current keyword targets.

How Do You Set Up Google Analytics 4 Integration?

Google Analytics 4 connects to Claude through community MCP servers. The GA4 MCP server enables queries about traffic, conversions, user behavior, and custom events using natural language.

GA4 MCP setup requires a service account with appropriate scopes:

Create a service account in Google Cloud Console
Grant Analytics Viewer permissions to your GA4 property
Download the JSON credentials file
Configure the MCP server with your property ID and credentials path

Once connected, Claude handles queries like:

“What was our top traffic source last week by session count?”
“Show conversion rate by device type for the past 30 days”
“Compare engagement metrics between organic and paid traffic”

GA4 integration works best when combined with the data plugin for SQL-based analysis. Export GA4 BigQuery data and query it through Claude’s data connector for advanced attribution modeling.

GA4 Gotchas for Marketing Teams

GA4 event parameter values have a 100 character limit. Custom dimensions must be registered before they appear in reports. DebugView has a 30-minute delay for some events. Enhanced measurement conflicts with GTM triggers if both track the same interactions.

Action item: Export your GA4 property ID and create a service account. Test the connection with “What was yesterday’s total sessions?” before building complex workflows.

How Do You Connect Microsoft Clarity to Cowork?

Microsoft Clarity provides a free, official MCP server for session recordings and behavioral analytics. The Clarity MCP server retrieves scroll depth, engagement time, traffic metrics, and session data filtered by device, browser, or country.

Add Clarity MCP in Claude Desktop:

{
  "mcpServers": {
    "@microsoft/clarity-mcp-server": {
      "command": "npx",
      "args": [
        "@microsoft/clarity-mcp-server",
        "--clarity_api_token=your-api-token-here"
      ]
    }
  }
}

Replace the token with your Clarity API token from your project settings. The server supports queries for 1-3 days of data, filtering by up to 3 dimensions.

Sample Clarity workflow for landing page optimization:

SYSTEM: You are a CRO analyst with Clarity access.

<context>
Landing page: {{PAGE_URL}}
Goal: Identify friction points reducing form completions
</context>

Fetch scroll depth and engagement time for the past 2 days, filtered by device type.

Output: Summary of mobile vs desktop behavior with 3 specific recommendations.

Microsoft released this MCP server in June 2025 as part of their AI-native analytics push. Marketing teams use it for pre-meeting analysis. Ask Claude to pull engagement data before stakeholder calls instead of logging into the Clarity dashboard.

Action item: Get your Clarity API token from Settings → API Access. Configure the MCP server and test with “Fetch scroll depth data for my site filtered by device type.”

How Do You Connect monday.com for Campaign Management?

monday.com offers an official MCP server available in Claude’s connector directory for Pro, Max, Team, and Enterprise plans. The monday MCP enables task creation, board queries, and status updates through natural language.

Connect monday.com to Claude:

Install the free monday MCP app from the monday marketplace (requires account admin)
In Claude, go to Connect your tools → Add connectors
Find monday.com and click Connect
Authorize via OAuth when redirected
Enable the connector in Search and tools menu

monday.com MCP capabilities include:

Project reporting and sprint summaries
Task creation from meeting notes
Cross-board visibility for campaign rollups
Status updates and owner assignments
CRM workflow automation for leads and deals

Team and Enterprise users need a two-step setup: organization-level configuration by Primary Owner, then individual user connection with OAuth.

monday.com Campaign Workflow Example

Create a campaign launch board automation:

SYSTEM: You are a marketing ops manager with monday.com access.

<context>
Campaign name: {{CAMPAIGN_NAME}}
Launch date: {{LAUNCH_DATE}}
Team: Marketing, Design, Content
</context>

Create a new board with groups for: Pre-Launch, Launch Week, Post-Launch.

Add items for: Creative brief, Landing page, Email sequence, Social posts, Paid ads, Analytics setup.

Assign owners based on team roles. Set due dates working backward from launch.

Output: Confirm board creation with link.

This replaces 30-45 minutes of manual board setup with a 2-minute prompt.

Action item: Have your monday.com admin install the MCP app. Connect your account and test with “List my monday.com boards.”

How Do You Connect HubSpot for CRM Automation?

HubSpot provides two MCP servers: a remote server for CRM data access and a developer server for CLI-based platform development. Both are in public beta with OAuth 2.1 support planned for later 2026.

Connect HubSpot remote MCP:

In Claude, navigate to Settings → Connectors
Click Browse connectors and select HubSpot
Click Add to your team (Super Admin required first time)
Authorize access to your HubSpot account
Select permission scopes for contacts, deals, companies, and engagement data

The HubSpot connector respects your user permissions. Individual sales reps see only deals they have access to. Claude cannot see data you cannot see in HubSpot.

HubSpot Marketing Workflows

Pipeline analysis prompt:

SYSTEM: You are a revenue analyst with HubSpot access.

Summarize all deals in the "Negotiation" stage with deal value > $10,000.

Include: Company name, deal owner, days in stage, next activity date.

Output: Table sorted by deal value. Flag any deals with no activity in 7+ days.

Lead enrichment workflow:

SYSTEM: You are a marketing ops specialist with HubSpot access.

<context>
Contact list: Enterprise prospects tagged "Q1 outreach"
Enrichment goal: Identify decision-maker job titles
</context>

Pull contacts from the specified list.

For each contact, check if job title contains VP, Director, or C-level.

Output: Segmented list with recommended next action per segment.

HubSpot MCP supports creating tasks, logging activities, and associating contacts with deals. Write permissions require explicit scope selection during setup.

Action item: Request Super Admin access to install HubSpot connector. Start with read-only scopes for contacts and deals. Expand permissions after testing.

How Do You Customize the Marketing Plugin for Your Team?

Plugin customization happens through skill files. Skills are markdown documents telling Claude how to handle specific workflows. Edit them to match your brand voice, approval processes, and campaign frameworks.

Customize skills by clicking the Customize button when viewing an installed plugin. Claude guides you through a Q&A workflow to generate tailored skill files. No manual editing required.

Brand Voice Skill Example

Create a skill file for brand-consistent content. The file contains frontmatter (name and description) followed by the actual instructions Claude follows:

---
name: Brand Voice Guide
description: Use when creating any marketing content.
---

# Brand Voice Instructions

## Tone
- Direct and confident, not salesy
- Write for busy executives who want facts
- Avoid jargon unless industry-standard

## Banned Phrases
Never use: "revolutionary", "game-changing", "synergy", "leverage"

## Approved Terminology
- Use "pipeline" not "sales funnel"
- Use "prospects" not "leads" for enterprise accounts
- Use "implementation" not "onboarding"

## Example Copy
Good: "Teams using this workflow report 40% faster campaign launches."
Bad: "Unlock revolutionary growth with our game-changing solution."

Claude references this skill automatically when you request content creation. The instructions become part of Claude’s context for every marketing content task.

Campaign Approval Workflow Skill

Document your approval process so Claude enforces it:

---
name: Campaign Approval Process
description: Use when launching any campaign or publishing content.
---

# Approval Workflow

Before any campaign goes live, complete these steps:

## Stage 1: Content Review
- Reviewer: Content Lead
- Checklist: Brand voice compliance, messaging accuracy, CTA clarity
- Timeline: 24 hours

## Stage 2: Legal Review (if claims made)
- Reviewer: Legal team
- Checklist: Compliance statements, disclaimers, competitor mentions
- Timeline: 48 hours

## Stage 3: Final Approval
- Reviewer: Marketing Director
- Checklist: Budget confirmed, tracking verified, launch timing approved

## Escalation
If approval stalls beyond timeline, notify VP Marketing via Slack.

Claude prompts you to complete each step before marking campaigns ready. Missing approvals trigger a reminder with the specific blocker.

Adding Custom Connectors

Edit .mcp.json to add connectors beyond the defaults:

{
  "mcpServers": {
    "semrush": {
      "url": "https://mcp.semrush.com/v1/mcp"
    },
    "clarity": {
      "command": "npx",
      "args": ["@microsoft/clarity-mcp-server", "--clarity_api_token=YOUR_TOKEN"]
    }
  }
}

Add as many MCP servers as your workflow requires. Each server authenticates independently.

Action item: Download the marketing plugin from GitHub (anthropics/knowledge-work-plugins/marketing) and review the skill files. Identify 3 skills to customize for your team’s processes.

What Enterprise Marketing Workflows Work Best with Cowork?

Enterprise marketing teams achieve the largest time savings on cross-platform reporting, competitive analysis, and campaign setup. These workflows combine multiple connectors and require context Claude previously lacked.

Weekly Performance Report Automation

Time saved: 3-4 hours weekly.

SYSTEM: You are a marketing analyst with access to GA4, HubSpot, and Semrush.

<context>
Reporting period: Last 7 days
Key metrics: Sessions, MQLs, Pipeline value, Organic keyword rankings
Audience: VP Marketing and CMO
</context>

MUST include:
1. Traffic summary with week-over-week comparison
2. Lead generation by source
3. Pipeline contribution from marketing campaigns
4. Top 5 keyword ranking changes
5. 3 action items for next week

Output: Executive summary format. 1 page max. Include specific numbers.

This prompt pulls data from three platforms, synthesizes insights, and formats for executive review.

Competitive Intelligence Brief

Time saved: 2-3 hours per competitor.

SYSTEM: You are a competitive intelligence analyst with Semrush and SimilarWeb access.

<context>
Competitor: {{COMPETITOR_NAME}}
Our company: {{YOUR_COMPANY}}
Focus areas: Traffic sources, keyword gaps, content strategy
</context>

Analyze:
1. Traffic volume and trend vs our site
2. Top 10 keywords they rank for that we don't
3. Content topics driving their organic growth
4. Paid search presence and estimated spend

Output: Battlecard format with "How to win" section.

Run this monthly for each competitor. Store outputs in Notion through the marketing plugin connector.

Campaign Launch Checklist

Time saved: 45 minutes per campaign.

SYSTEM: You are a marketing ops manager with monday.com and HubSpot access.

<context>
Campaign: {{CAMPAIGN_NAME}}
Type: {{CAMPAIGN_TYPE}}
Launch date: {{LAUNCH_DATE}}
</context>

Create campaign infrastructure:

1. monday.com board with pre-launch tasks and assignments
2. HubSpot campaign record for attribution tracking
3. UTM parameter documentation
4. QA checklist for tracking verification

Confirm each item created with links.

This workflow replaces manual setup across two platforms with a single prompt execution.

Action item: Map your current reporting workflow. Identify which data sources feed each report. Build a prompt combining those sources through Cowork connectors.

What Are the Limitations and Gotchas?

Cowork plugins run as a research preview with specific constraints. Understanding these limitations prevents frustration during implementation.

Current Limitations

Plugins save locally to your machine. Organization-wide plugin sharing comes in a future update. No project support exists inside Cowork sessions. Claude does not retain memory between Cowork sessions. Sessions cannot be shared with other team members.

Cowork runs only in Claude Desktop for macOS. Windows and Linux support is not announced. The app must remain open for sessions to continue. Closing the app ends the session.

MCP Server Gotchas

Each MCP tool has a schema definition consuming context tokens. Loading too many MCP servers bloats context and reduces Claude’s effective working memory. The GitHub and Linear official MCPs consume 17.1K tokens combined (8.5% of context).

Solutions:

Enable only connectors you need for the current session
Use Tool Search mode when MCP tools exceed 10% of context
Build separate plugins for different workflow types instead of one mega-plugin

Data and Security Considerations

Cowork activity does not appear in Audit Logs, Compliance API, or Data Exports. Do not use Cowork for regulated workloads requiring audit trails.

Conversation history stores locally on your computer. This data falls outside Anthropic’s standard retention policies. Review your organization’s data handling requirements before enabling Cowork.

Action item: Document which workflows require audit trails. Keep those in standard Claude or your approved automation tools. Use Cowork for non-regulated operational workflows.

How Do You Build a Custom Marketing Plugin from Scratch?

Use the Cowork Plugin Management plugin to create custom plugins. This plugin walks you through skill creation, connector configuration, and command setup without manual file editing.

Step 1: Define Your Plugin Scope

Start with a specific job function or workflow. Avoid building an everything plugin. Examples:

Paid media operations plugin
Content marketing plugin
Demand generation plugin
Marketing analytics plugin

Narrow scope produces better Claude behavior than broad generic plugins.

Step 2: Create Your Plugin

In Cowork, invoke:

/cowork-plugin-management:create

Claude prompts you for:

Plugin name and description
Target role or function
Primary tools and data sources
Key workflows to support

The plugin generates a folder structure with placeholder files.

Step 3: Add Connectors

Edit .mcp.json to add your tool connections:

{
  "mcpServers": {
    "hubspot": {
      "url": "https://mcp.hubspot.com"
    },
    "ga4": {
      "command": "npx",
      "args": ["ga4-mcp-server", "--credentials=path/to/credentials.json"]
    }
  }
}

Step 4: Write Skills

Create skill files for each major workflow. Skills contain frontmatter plus detailed instructions:

---
name: Campaign Performance Analysis
description: Use when user asks about campaign metrics or performance.
---

# Campaign Performance Analysis

When analyzing campaign performance, follow this structure:

## Required Data Points
- Traffic by source (GA4)
- Conversion rate by channel
- Cost per acquisition
- Pipeline contribution (HubSpot)

## Analysis Framework
1. Compare to previous period (default: week-over-week)
2. Identify top and bottom performers
3. Flag anomalies (>20% deviation from average)

## Output Format
- Executive summary (3 sentences max)
- Metrics table with trends
- 3 recommended actions

The body tells Claude exactly how to execute the workflow. More detailed skills produce more consistent outputs.

Step 5: Add Slash Commands

Slash commands let you trigger specific workflows with a single keystroke. Type / in Cowork to see available commands from your installed plugins. Commands appear with the format /pluginname:commandname.

Example slash commands from the marketing plugin:

/marketing:campaign-brief generates a campaign brief template
/marketing:competitor-analysis pulls competitive intelligence
/marketing:content-calendar creates a content planning board
/marketing:performance-report generates weekly metrics summary

Invoke a command with arguments by adding them after the command:

/marketing:performance-report last 14 days

Commands execute pre-defined workflows without requiring detailed prompts. They standardize outputs across your team since everyone uses the same underlying skill logic.

Action item: Identify your team’s most repetitive marketing task. Build a custom plugin with one skill and one command addressing this workflow. Test for one week before expanding.

What Does a Complete Marketing Stack Integration Look Like?

A fully integrated marketing stack in Cowork combines 5-7 connectors covering CRM, analytics, SEO, project management, and communication. Here is a reference architecture for enterprise B2B marketing teams.

Reference Marketing Stack

Core connectors:

HubSpot: CRM, marketing automation, deal pipeline
GA4: Web analytics, conversion tracking, attribution
Semrush: Keyword research, competitor analysis, rank tracking
monday.com: Campaign management, task tracking, resource allocation
Slack: Team communication, alerts, approvals

Supporting connectors:

Microsoft Clarity: User behavior, session recordings, heatmaps
Notion: Documentation, knowledge base, SOPs

Data Flow Architecture

Semrush → Keyword insights → Content planning (Notion)
                                    ↓
                            Content production
                                    ↓
GA4 ← Traffic/conversion ← Published content
 ↓
HubSpot ← Lead capture ← Conversion events
 ↓
monday.com ← Campaign tasks ← Performance triggers
 ↓
Slack ← Notifications ← Status updates

Claude orchestrates data flow between platforms. Ask for insights spanning multiple sources without manual exports or data wrangling.

Sample Cross-Platform Prompt

SYSTEM: You are a marketing strategist with access to HubSpot, GA4, Semrush, and monday.com.

<context>
Goal: Plan next quarter's content calendar
Constraints: 2 blog posts per week, 1 pillar page per month
Budget: 20 hours content production weekly
</context>

Execute this workflow:

1. Pull top 50 converting keywords from GA4 (last 90 days)
2. Cross-reference with Semrush keyword difficulty
3. Identify gaps versus top 3 competitors
4. Check HubSpot for existing content on target keywords
5. Create monday.com board with prioritized content calendar

Output: Content calendar board with keyword targets, due dates, and estimated traffic potential.

This prompt spans 4 platforms and produces a ready-to-execute content plan.

Action item: Map your current marketing tools to Cowork connector availability. Identify gaps where community MCP servers exist. Build your target architecture before implementing.

Final Takeaways

Cowork plugins transform Claude from a generic assistant into a marketing operations specialist connected to your actual tools and data.

The marketing plugin ships with 9 connectors but requires customization to match your brand voice, processes, and approval workflows.

MCP servers for Semrush, GA4, Microsoft Clarity, monday.com, and HubSpot connect through official or community implementations with minimal configuration.

Enterprise workflows combining multiple data sources save 15-20 hours weekly on reporting, competitive analysis, and campaign setup.

Plugin customization requires no code. Edit markdown skill files to teach Claude your specific terminology and workflows. Start with one high-impact workflow before expanding.