# LinkedIn MCP Server Setup

## Prerequisites

1. **Python 3.12+** on your host machine
2. **LinkedIn account** with valid credentials
3. **Create LinkedIn profile locally** (one-time setup)

## Step 1: Install LinkedIn MCP (Host Machine)

Run this on your **host machine** (not in Docker):

```bash
pip install linkedin-scraper-mcp
```

## Step 2: Create LinkedIn Profile (One-time Authentication)

Run this command to authenticate and create your profile:

```bash
linkedin-mcp-server --login
```

Or use the alternate command:

```bash
linkedin-scraper-mcp --login
```

This will:
- Open a browser window
- Ask you to log into LinkedIn
- Handle captcha challenges (manual solving required)
- Save your session at `~/.linkedin-mcp/profile/`

**Important:** Only one active LinkedIn session is supported. If you need to re-authenticate:
```bash
rm -rf ~/.linkedin-mcp/profile
linkedin-mcp-server --login
```

## Step 3: Build Docker Image

```bash
cd /mnt/NVME/Docker/mcp-gateway
docker compose build linkedin-mcp
```

## Step 4: Start the Service

```bash
docker compose up linkedin-mcp
```

The service is already configured in your `docker-compose.yml` with:
- Python 3.12 slim image
- LinkedIn profile mounted from `~/.linkedin-mcp/profile`
- LOG_LEVEL environment variable from `.env`
- Automatic restart

## Configuration Options

### Environment Variables (in .env)

```env
# LinkedIn MCP Server
LINKEDIN_LOG_LEVEL=INFO
LINKEDIN_TIMEOUT=10000
```

### Available Log Levels

- `DEBUG` - Detailed logging
- `INFO` - Standard logging (recommended)
- `WARNING` - Only warnings and errors
- `ERROR` - Only errors

## Docker Integration

The Dockerfile:
- Uses `python:3.12-slim` as base image
- Clones the GitHub repository
- Installs dependencies with pip
- Mounts profile from host at `~/.linkedin-mcp/profile:/root/.linkedin-mcp/profile`

## Troubleshooting

### Profile Not Found
```bash
# Verify profile exists on host
ls ~/.linkedin-mcp/profile

# Create new profile if missing
linkedin-mcp-server --login
```

### Build Fails - Git Clone Issues
- Ensure you have internet access
- Check git is installed: `git --version`
- Try manual clone first: `git clone https://github.com/stickerdaniel/linkedin-mcp-server.git`

### Container Won't Start
Check the logs:
```bash
docker logs mcp-linkedin

# Restart container
docker compose restart mcp-linkedin
```

### Session Expired
LinkedIn sessions may expire. To refresh:
```bash
# Delete old profile
rm -rf ~/.linkedin-mcp/profile

# Create new profile
linkedin-mcp-server --login

# Restart container
docker compose restart mcp-linkedin
```

### Captcha Loop
If you keep hitting captchas:
1. Create profile manually on your machine first
2. Mount that profile into Docker
3. Run with `--no-headless` temporarily to debug

## Integration with Gateway Proxy

The gateway proxy (`gateway_proxy.py`) should have:

```python
MCP_BACKEND_LINKEDIN=http://mcp-linkedin:3000/mcp
```

This is already configured in the docker-compose.yml.

## Verification

Test the connection:

```bash
# Check if container is running
docker ps | grep mcp-linkedin

# View logs
docker logs mcp-linkedin

# Check gateway can reach it
docker exec mcp-gateway curl http://mcp-linkedin:3000/mcp/tools
```

## Entry Point

The container runs:
```
linkedin-mcp-server --log-level INFO
```

This starts the MCP server that communicates with the gateway proxy via HTTP.

## Performance Notes

- First run may be slower as Patchright downloads browser dependencies
- Profile authentication takes 30-60 seconds
- Subsequent tool calls should respond in 1-5 seconds
- Browser automation is CPU-intensive; allocate resources accordingly