linkedin-mcp-server

A full-featured LinkedIn MCP server for Claude. Post content, schedule publications, generate AI images, manage comments and reactions — all from your AI assistant.

What It Does

Connect your LinkedIn account to Claude and manage your professional presence with natural language:

"Schedule a thought leadership post about AI automation for Monday 9am with a Gemini-generated banner image"

"Reply to the latest comments on my posts"

"Create a case study post about our MCP server launch using the case-study template"

25 MCP Tools

Post Management — Create, edit, delete, repost, read your posts Scheduling — Schedule posts for future publication with automatic background publishing Media — Upload images and videos from local files or URLs AI Images — Generate images with Google Gemini Imagen 4 and post them directly Comments & Reactions — Comment, reply, react (LIKE, CELEBRATE, SUPPORT, LOVE, INSIGHTFUL, FUNNY) Content Templates — 12 built-in templates (Thought Leadership, Case Study, Viral Trend, Lead Magnet, and more) plus custom ones Brand Voice — Configure tone, emoji style, hashtag strategy, post structure, and LinkedIn algorithm settings Algorithm Guidelines — Built-in LinkedIn algorithm knowledge: ranking factors, copywriting rules, hooks, CTAs, timing, checklist

Quick Start

1. Create a LinkedIn App

1. Go to LinkedIn Developer Portal
2. Create a new app
3. Under Products, enable:
   • "Share on LinkedIn"
   • "Sign In with LinkedIn using OpenID Connect"
4. Under Auth, add redirect URL: http://localhost:8585/callback
5. Note your Client ID and Client Secret

2. Configure for Claude Code

claude mcp add --scope user linkedin npx -- -y linkedin-mcp-server \
  -e LINKEDIN_CLIENT_ID=your_client_id \
  -e LINKEDIN_CLIENT_SECRET=your_client_secret \
  -e GEMINI_API_KEY=your_gemini_key

3. Configure for Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "linkedin": {
      "command": "npx",
      "args": ["-y", "linkedin-mcp-server"],
      "env": {
        "LINKEDIN_CLIENT_ID": "your_client_id",
        "LINKEDIN_CLIENT_SECRET": "your_client_secret",
        "GEMINI_API_KEY": "your_gemini_key"
      }
    }
  }
}

4. Authenticate

Ask Claude:

"Connect my LinkedIn account"

Claude will generate an authorization URL. Open it in your browser, approve access, and tokens are saved automatically (valid for 60 days with auto-refresh).

5. Start Posting!

"Post on LinkedIn: Just shipped a LinkedIn MCP server with 24 tools. The future of AI-powered content management is here. #MCP #AI #LinkedIn"

Tools Reference

Tool	Description
linkedin_auth_start	Start OAuth 2.0 authorization flow
linkedin_auth_status	Check auth status and token expiry
linkedin_profile_me	Get your profile info and person URN
linkedin_brand_voice	Get/set brand voice preferences
linkedin_post_create	Create a post (text, image, video, article)
linkedin_post_update	Edit an existing post
linkedin_post_delete	Delete a post
linkedin_post_repost	Reshare with optional commentary
linkedin_post_get	Get post details
linkedin_posts_list	List your recent posts
linkedin_comment_create	Add a comment or reply
linkedin_comments_list	List post comments
linkedin_comment_delete	Delete your comment
linkedin_reaction_add	React to a post or comment
linkedin_reaction_remove	Remove your reaction
linkedin_media_upload	Upload image/video from file or URL
linkedin_gemini_image	Generate AI image with Gemini Imagen 4
linkedin_schedule_create	Schedule a post for future publication
linkedin_schedule_list	View all scheduled posts
linkedin_schedule_update	Modify a scheduled post
linkedin_schedule_cancel	Cancel a scheduled post
linkedin_template_list	Browse content templates
linkedin_template_get	View template details and variables
linkedin_template_save	Create or update a custom template
linkedin_guidelines	LinkedIn algorithm strategy and best practices

Post Scheduling

The server includes a built-in scheduler that runs alongside the MCP server:

1. Schedule a post with a target datetime (ISO 8601)
2. Background daemon checks every 30 seconds for due posts
3. Posts are published automatically at the scheduled time
4. Failed posts retry up to 3 times with 5-minute intervals
5. On startup, overdue posts (up to 24h) are caught up automatically

You can also run the scheduler as a standalone daemon:

npx linkedin-mcp-server --scheduler

AI Image Generation

Generate images with Google Gemini Imagen 4 and post them directly to LinkedIn.

Get a free API key at Google AI Studio and set GEMINI_API_KEY.

Scheduled posts can auto-generate images at publish time using the gemini_prompt parameter — the image is created fresh right before publishing.

Content Templates

12 built-in templates for consistent, high-quality posts:

Template	Structure
Thought Leadership	Hook → Insights → CTA
Thought Leadership (PL)	Hook → Kontekst → Insighty → CTA
Case Study	Problem → Solution → Results
Announcement	Headline → Details → Features → CTA
Engagement Hook	Question → Context → Options
Behind the Scenes	Intro → Story → Lesson
Lesson Learned	Confession → Mistake → Lesson → Advice
Carousel Text	Title → Slides → CTA
Carousel Edu	Hook → Steps → Summary → CTA
Viral Trend	Hook → Trend → Stat → Your Take → CTA
Community Question	Hook → Context → Question → CTA
Lead Magnet	Hook → Problem → What You Get → CTA

Create your own templates with {{variable}} syntax using linkedin_template_save.

Manual Token Mode

If you already have a LinkedIn access token (e.g., from n8n or another tool):

{
  "env": {
    "LINKEDIN_ACCESS_TOKEN": "AQVxxx...",
    "LINKEDIN_PERSON_URN": "urn:li:person:your_id"
  }
}

This skips OAuth entirely. Useful for CI/CD or existing integrations.

Environment Variables

Variable	Required	Description
LINKEDIN_CLIENT_ID	Yes*	LinkedIn app Client ID
LINKEDIN_CLIENT_SECRET	Yes*	LinkedIn app Client Secret
LINKEDIN_ACCESS_TOKEN	Alt*	Manual token (skips OAuth)
LINKEDIN_PERSON_URN	With manual	urn:li:person:xxx
GEMINI_API_KEY	No	For AI image generation
LINKEDIN_CALLBACK_PORT	No	OAuth port (default: 8585)
LINKEDIN_DATA_DIR	No	Data dir (default: ~/.linkedin-mcp)
LINKEDIN_API_VERSION	No	API version (default: 202503)

*Either Client ID + Secret OR Access Token is required.

Development

git clone https://github.com/gacabartosz/linkedin-mcp-server.git
cd linkedin-mcp-server
npm install
npm run dev   # Run with tsx (hot reload)
npm run build # Compile TypeScript

Built With

This project was built using the Model Context Protocol and follows patterns from the Anthropic MCP Builder Skill.

Architecture and code patterns were informed by studying existing LinkedIn MCP projects — thanks to the community for the inspiration:

• stickerdaniel/linkedin-mcp-server — great work on LinkedIn profile scraping and the Patchright-based session approach
• rugvedp/linkedin-mcp — useful reference for post data structures
• felipfr/linkedin-mcpserver — TypeScript MCP patterns and OAuth architecture ideas

Development was accelerated by Ralph — an autonomous AI dev loop that runs Claude Code in iterative cycles for automated implementation, testing, and code polish.

Author

Bartosz Gaca — bartoszgaca.pl — kontakt@bartoszgaca.pl

License

MIT