linkedin-prospection-mcp
# LinkedIn Prospection MCP
MCP server for LinkedIn prospection automation — find leads, score (fit+intent+urgency), qualify, personalize messages, run full pipeline, manage sales funnel.
## Installation
```bash
npx linkedin-prospection-mcp
```
Or install globally:
```bash
npm install -g linkedin-prospection-mcp
```
## Configuration
### Claude Desktop
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"linkedin-prospection": {
"command": "npx",
"args": ["-y", "linkedin-prospection-mcp"],
"env": {
"PROSPECTION_DIR": "/path/to/your/prospection/scripts"
}
}
}
}
```
### Claude Code
Add to `.claude/settings.json`:
```json
{
"mcpServers": {
"linkedin-prospection": {
"command": "npx",
"args": ["-y", "linkedin-prospection-mcp"],
"env": {
"PROSPECTION_DIR": "/path/to/your/prospection/scripts"
}
}
}
}
```
## Tools
| Tool | Description |
|------|-------------|
| `find_leads` | Search LinkedIn for leads matching burnout/stress signals |
| `score_lead` | Score a single lead (fit 0-30 + intent 0-40 + urgency 0-30 = /100) |
| `qualify_leads` | Batch qualify leads with P1-P4 priority classification |
| `personalize_message` | Generate personalized invitation notes and DM sequences |
| `run_pipeline` | Run the full daily prospection pipeline |
| `get_pipeline_status` | Get current pipeline status and daily log |
| `manage_lead` | Update lead status in the pipeline |
## Resources
- `linkedin-prospection://daily-log` — Today's prospection log
- `linkedin-prospection://leads` — All discovered leads
## Prompts
- `daily_prospection` — Guided daily prospection workflow (full or quick mode)
## Scoring Engine
Leads are scored on three axes:
- **Fit (0-30)**: Role match, industry risk, seniority
- **Intent (0-40)**: Burnout keywords, stress signals, help-seeking language
- **Urgency (0-30)**: Recency, crisis indicators, explicit requests
Priority classification:
- **P1-hot** (70+): Immediate outreach
- **P2-warm** (50-69): Nurture sequence
- **P3-nurture** (30-49): Long-term follow-up
- **P4-cold** (<30): Archive
## Environment Variables
| Variable | Description | Default |
|----------|-------------|---------|
| `PROSPECTION_DIR` | Path to prospection scripts directory | `./lib/prospection` |
## License
MIT
TDQS
Scored across 7 tools
Tools are generally distinct, though `qualify_leads` and `score_lead` serve similar scoring functions (batch vs. individual). The `personalize_message` tool generates content without sending, which could be confused with messaging actions embedded in `run_pipeline`, but descriptions clarify this adequately.
All tools follow a consistent verb_noun pattern in snake_case (find_leads, manage_lead, run_pipeline, etc.). The vocabulary is domain-appropriate and predictable throughout the set.
Seven tools is appropriate for a LinkedIn prospecting domain, falling within the ideal range. Slight redundancy exists between individual (`score_lead`) and batch (`qualify_leads`) scoring operations, but each earns its place for different workflow patterns.
Notable gaps exist: no `get_lead` or `list_leads` to read individual pipeline entries, and no standalone `send_invitation` or `send_dm` tool (sending only available through the automated `run_pipeline`). Agents can work around these using `get_pipeline_status` and `run_pipeline`, but granular control is missing.