LinkedIn MCP Pro¶

LinkedIn automation for AI agents — 54 MCP tools, ban-safety gates, full web dashboard, MIT licensed.

LinkedIn MCP Pro is a Model Context Protocol server that gives any AI agent a complete LinkedIn workflow: compose, schedule, engage, search jobs, track applications, monitor account health — all behind safety gates that keep your account out of ban territory.

It ships with a unified web dashboard (15 pages, Linear + LinkedIn design system) and 10 CLI commands for headless operation.

✨ Features¶

For AI agents (MCP)¶

54 tools across drafts, scheduling, engagement, jobs, analytics, safety, multi-account
stdio MCP server — works with Claude Desktop, Cursor, Cline, Windsurf, Zed, Continue, Open WebUI, LibreChat, LobeChat
Ban-safety gates built into every write operation
Multi-account with per-account persona + rate limits

For humans (Web UI)¶

15-page dashboard with persistent sidebar + topbar shell
Linear dark mode (#5e6ad2 accent) + LinkedIn content cards (#0a66c2 brand)
Jobs module — CV upload, profile wizard, search, match scoring, cover letter, application tracker
Real-time analytics — engagement heatmap, best-time-to-post, ab testing results
Mobile responsive with collapsible sidebar

For ops (CLI)¶

linkedin-mcp-web — start the dashboard
linkedin-mcp-install — wire any MCP host in one command
linkedin-mcp-templates — manage post templates
linkedin-mcp-schedule — cron + queue management
linkedin-mcp-stats / linkedin-mcp-health — diagnostics
linkedin-mcp-login — headless cookie import
linkedin-mcp-deadman — watchdog for stalled automations
linkedin-mcp-analytics — CSV / JSON export
linkedin-mcp-pro — stdio MCP server

🚀 Quick start¶

pipx (recommended)pipFrom source

pipx install git+https://github.com/horizonbymuneeb/linkedin-mcp-pro.git@v2.3.9
linkedin-mcp-web --host 0.0.0.0 --port 8080

Open http://localhost:8080

pip install --user git+https://github.com/horizonbymuneeb/linkedin-mcp-pro.git@v2.3.9
linkedin-mcp-web

git clone https://github.com/horizonbymuneeb/linkedin-mcp-pro.git
cd linkedin-mcp-pro
pip install -e .
linkedin-mcp-web

Then point your AI agent at the MCP server:

{
  "mcpServers": {
    "linkedin-mcp-pro": {
      "command": "linkedin-mcp-pro",
      "args": ["serve"]
    }
  }
}

First-run checklist

Open the dashboard at http://localhost:8080/connect
Log into LinkedIn (browser-driven cookie import — no password leaves your machine)
Visit /jobs to upload your CV (or skip — drafts/scheduling work without it)
Wire your agent via /install
Compose your first draft via /drafts

🏗 Architecture at a glance¶

┌─────────────────┐     stdio JSON-RPC      ┌──────────────────┐
│  AI Agent Host  │ ◄─────────────────────► │  linkedin-mcp-pro │
│ (Claude/Cursor) │                          │   (FastAPI+CLI)   │
└─────────────────┘                          └────────┬───────────┘
                                                      │
                                  ┌───────────────────┼───────────────────┐
                                  ▼                   ▼                   ▼
                          ┌──────────────┐    ┌──────────────┐    ┌──────────────┐
                          │ Web Dashboard│    │ SQLite DB    │    │ Browser Auto │
                          │  (15 pages)  │    │  (cookies,   │    │  (Playwright)│
                          │              │    │   drafts,    │    │              │
                          │              │    │   schedules) │    │              │
                          └──────────────┘    └──────────────┘    └──────────────┘

Read the full architecture doc.

📚 Documentation¶

Getting started

Install · Quickstart · Architecture · Agent setup
Guides

Dashboard tour · Jobs module · Drafts & templates · Scheduling · Engagement · Multi-account
Reference

CLI commands · REST API · MCP tools (54)
Operations

Safety system · Proxy setup · Troubleshooting · Roadmap

🛡 Safety guarantees¶

Every write path runs through a ban-safety gate that checks:

Daily action caps (configurable per account)
Velocity windows (no more than N actions per hour)
Content patterns (LinkedIn spam triggers blocked)
Duplicate detection (no identical posts within 7 days)
Rate-limit backoff when LinkedIn throttles

The gate is always on — there is no --unsafe flag. See safety.md for the full policy.

🤝 Contributing¶

We welcome PRs! See contributing.md for setup, style guide, and the test-first workflow.