linkedin-mcp-pro
README.md
# LinkedIn MCP Pro
> **LinkedIn automation for AI agents — 54 MCP tools, ban-safety gates, full web dashboard, MIT licensed.**
[](https://github.com/horizonbymuneeb/linkedin-mcp-pro)
[](https://github.com/horizonbymuneeb/linkedin-mcp-pro/releases)
[](https://www.python.org/)
[](LICENSE)
[](https://horizonbymuneeb.github.io/linkedin-mcp-pro/)
**[📚 Full documentation](https://horizonbymuneeb.github.io/linkedin-mcp-pro/)** · **[Quickstart](https://horizonbymuneeb.github.io/linkedin-mcp-pro/getting-started/quickstart/)** · **[API reference](https://horizonbymuneeb.github.io/linkedin-mcp-pro/reference/api/)** · **[MCP tools](https://horizonbymuneeb.github.io/linkedin-mcp-pro/reference/mcp-tools/)**
---
## What is it?
**LinkedIn MCP Pro** is a [Model Context Protocol](https://modelcontextprotocol.io/) server that gives any AI agent a complete LinkedIn workflow:
- ✍️ **Compose** posts (manual / AI / template)
- 📅 **Schedule** posts via cron-like queue
- 💬 **Engage** with comments, DMs, mentions
- 🔍 **Search** jobs with match scoring + auto cover letter
- 📊 **Analytics** with engagement heatmaps + A/B testing
- 🛡 **Ban-safety gates** on every write path
It ships with a unified **web dashboard** (15 pages) and **10 CLI commands**.
---
## ⚡ Quick start
```bash
# Install
pipx install git+https://github.com/horizonbymuneeb/linkedin-mcp-pro.git@v2.3.2
# Start dashboard
linkedin-mcp-web --host 0.0.0.0 --port 8080
# Or run as stdio MCP server for AI agents
linkedin-mcp-pro serve
```
Open <http://localhost:8080> for the dashboard, or wire your agent:
```bash
linkedin-mcp-install add claude-desktop # or cursor, cline, windsurf, zed, ...
```
Full guide: **[Quickstart](https://horizonbymuneeb.github.io/linkedin-mcp-pro/getting-started/quickstart/)**
---
## 📊 What's inside
| Surface | Count |
|---------|-------|
| **MCP tools** | 54 (across 10 families) |
| **REST endpoints** | 60+ |
| **CLI commands** | 10 |
| **Dashboard pages** | 15 (unified Linear+LinkedIn design system) |
| **Tests** | 721 passing |
---
## 🖼 Dashboard preview
The dashboard at `/` is a unified shell — same sidebar, topbar, design tokens across every page:
- **Workspace**: Home, Drafts, Schedule, Engage, **Jobs**, Analytics
- **Account**: Connect, Cookies, Profile
- **Configure**: LLM, Safety, Audit, Install, Settings, Templates
Design system: **Linear dark mode** (`#5e6ad2` accent) + **LinkedIn content cards** (`#0a66c2` brand).
See [Dashboard tour](https://horizonbymuneeb.github.io/linkedin-mcp-pro/guides/dashboard/) for a walkthrough.
---
## 🤖 Agent support
Works with any MCP host:
- Claude Desktop (macOS / Linux / Windows)
- Claude Code
- Cursor
- Cline (VS Code)
- Continue.dev
- Windsurf
- Zed
- Open WebUI
- LibreChat
- LobeChat
Full setup: [Agent setup guide](https://horizonbymuneeb.github.io/linkedin-mcp-pro/getting-started/agents/)
---
## 🛡 Safety
Every write path runs through a **ban-safety gate**:
- Daily action caps (per-account configurable)
- Velocity windows (no more than N actions per hour)
- Content pattern checks (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](https://horizonbymuneeb.github.io/linkedin-mcp-pro/operations/safety/).
---
## 📦 What ships
```
linkedin-mcp-pro/
├── linkedin_mcp/ # Core package
│ ├── static/ # 16 HTML pages + unified shell
│ ├── jobs/ # Jobs module (CV, search, apply, tracker)
│ ├── tools/ # 17 tool modules
│ ├── cli*.py # 10 CLI entry points
│ └── web.py # FastAPI server
├── tests/ # 721 tests
├── scripts/
│ └── e2e_test.py # Full E2E smoke test (64 checks)
├── docs/ # MkDocs Material documentation
├── mkdocs.yml # Docs build config
├── install.sh / install.ps1 # One-line installers
└── README.md # This file
```
---
## 🧪 Development
```bash
git clone https://github.com/horizonbymuneeb/linkedin-mcp-pro.git
cd linkedin-mcp-pro
pip install -e .
pytest tests/ # 721 tests
python scripts/e2e_test.py # E2E smoke (64 checks)
# Build docs locally
pip install mkdocs mkdocs-material pymdown-extensions
mkdocs serve # → http://127.0.0.1:8000
```
---
## 📄 License
MIT © [horizonbymuneeb](https://github.com/horizonbymuneeb)
See [LICENSE](LICENSE).
---
## 🔗 Links
- 📚 [Documentation](https://horizonbymuneeb.github.io/linkedin-mcp-pro/)
- 🐛 [Issue tracker](https://github.com/horizonbymuneeb/linkedin-mcp-pro/issues)
- 💬 [Discussions](https://github.com/horizonbymuneeb/linkedin-mcp-pro/discussions)
- 🚀 [Releases](https://github.com/horizonbymuneeb/linkedin-mcp-pro/releases)TDQS
A3.6/5.0
Scored across 58 tools
Disambiguation5/5
Each tool targets a specific action or resource, with clear boundaries. For example, auto_comment_by_keyword, auto_connect_by_criteria, and auto_like_by_keyword are distinct automation actions. Even similar tools like get_daily_stats and get_quota_usage have different scopes (daily vs raw). No overlapping purposes.
Naming Consistency5/5
All tools follow a consistent verb_noun pattern using underscore_case, e.g., accept_invitation, create_post, search_people. Compound verbs like auto_comment_by_keyword are predictable and clear. No mixing of conventions.
Tool Count4/5
At 58 tools, the count is high but appropriate for a 'pro' server covering posting, messaging, automation, scheduling, templates, analytics, safety, and LLM management. Each tool serves a specific need, though the surface is extensive.
Completeness5/5
The tool set covers the full lifecycle of LinkedIn automation: connection management, messaging, content creation, scheduling, feed interaction, analytics, safety, and configuration. Minor gaps like profile editing are outside the server's stated purpose.
Maintenance
ActivityInactive
ResponsivenessNo issues