MCP Servers

A collection of Model Context Protocol servers, templates, tools and more.

G
Gtm MCP
by @petrNikolaev1
GitHub Repo(1 stars)

MCP server by petrNikolaev1

Created 4/2/2026
Updated about 7 hours ago
README
View Source
Repository documentation and setup instructions

MCP LeadGen — AI-Powered Lead Generation Platform

An open-source MCP (Model Context Protocol) server that enables AI agents (Claude, GPT, etc.) to manage B2B lead generation pipelines end-to-end.

Want to skip the setup? Use the hosted version at gtm-mcp.com — sign up and start generating leads in minutes, no infrastructure required.

What It Does

Connect your AI assistant to this MCP server and it can:

  • Search Apollo for target companies and contacts
  • Scrape and analyze company websites
  • Classify leads using AI (OpenAI / Gemini / Anthropic)
  • Enrich contacts with verified emails
  • Push qualified leads to SmartLead email campaigns
  • Automate LinkedIn outreach via GetSales
  • Monitor campaign replies and categorize responses
  • Learn from operator corrections — the system tracks approved/dismissed reply drafts to improve over time

Example Queries

Just tell your AI agent what you want in plain English:

Single-segment campaign

easystaff.io
IT consulting companies in Miami

Creates a project from the website, then gathers IT consulting companies in Miami via Apollo, classifies them, extracts contacts, and pushes to a SmartLead campaign as DRAFT.

Multi-segment campaign

easystaff.io
IT consulting companies in Miami
Video production in London

Creates separate pipelines per segment. Each runs independently with its own keywords, filters, and campaign.

From a website

https://thefashionpeople.com/
Fashion brands in Italy

Scrapes the website to understand the offer, auto-generates 20+ Apollo keywords for fashion brands, gathers and classifies companies in Italy.

Extended geo for faster KPI

IT consulting companies in US
Video production in UK

Broader geo = more companies available = KPI (100 contacts) reached faster.

Check your data

How much total contacts in apollo?

Multi-segment in one message

Find IT consulting companies in Miami and video production companies in London

The system parses this into 2 separate segments and confirms before launching.

Launch from a strategy document

Write your outreach strategy in a file (offer, ICP, segments, sequence style, sender info) and tell your AI agent:

Launch outreach-plan.md

The agent reads your file and passes it to create_project with document_text. The system automatically extracts everything — offer, target roles, segments, email sequence style, campaign settings — and creates a fully configured project. You just confirm the offer and it runs.

Example outreach-plan.md:

# Company: EasyStaff
## Offer
Global payroll & contractor payments for companies hiring remote teams.
## ICP
- IT outsourcing companies, 50-500 employees
- SaaS companies with distributed teams
- Digital agencies scaling internationally
## Segments
- IT consulting companies in US
- SaaS companies in DACH region
- Digital agencies in UK
## Sender
Name: Alex, Position: Partnership Manager, Company: EasyStaff
## Sequence style
Casual, short, question-based. 4 emails. Reference their tech stack.

This replaces manual project setup — one file drives the entire campaign creation flow.

Pipeline Flow

The strict flow ensures no credits are wasted:

1. create_project (website → offer extraction)     → WAIT for approval
2. confirm_offer                                    → WAIT
3. align_email_accounts (select SmartLead senders)  → WAIT
4. tam_gather PREVIEW (probe Apollo, show strategy) → WAIT for "Proceed?"
5. tam_gather CONFIRM → pipeline runs autonomously:
   gather → scrape → classify → extract people → push to SmartLead DRAFT

Default KPIs: 100 contacts, 3 per company. Pipeline stops automatically when KPI is hit.

Architecture

AI Agent (Claude/GPT) ──MCP Protocol──► Backend (FastAPI)
                                            │
                                    ┌───────┼───────┐
                                    ▼       ▼       ▼
                                PostgreSQL Redis  External APIs
                                (pgvector)        (Apollo, SmartLead,
                                                   OpenAI, GetSales)
                                            │
Frontend (React) ◄──────────────────────────┘
  • Backend: FastAPI + async SQLAlchemy + MCP SDK (SSE transport)
  • Frontend: React 19 + TypeScript + Vite + Tailwind + AG Grid
  • Database: PostgreSQL with pgvector extension
  • Cache: Redis for rate limiting and session data

Quick Start

Prerequisites

  • Docker and Docker Compose

1. Clone and configure

git clone <repo-url> mcp-leadgen
cd mcp-leadgen
cp .env.example .env
# Edit .env — set POSTGRES_PASSWORD and add your API keys

2. Start services

docker compose up --build

3. Access the UI

Open http://localhost:3000 — create an account, then configure your API keys on the Setup page.

4. Connect your AI agent

The MCP SSE endpoint is available at:

http://localhost:8002/mcp/sse?token=YOUR_MCP_TOKEN

Get your token from the Setup page after signing up.

API Keys

Configure these on the Setup page (per-user, encrypted at rest):

| Service | Purpose | Required | |---------|---------|----------| | OpenAI | Lead analysis & classification | Yes (or Gemini) | | Apollo | Company/people search | Yes | | SmartLead | Email campaign management | For outreach | | Apify | Website scraping (proxy) | For scraping | | GetSales | LinkedIn outreach automation | Optional |

All MCP Tools

Account (3)

| Tool | Description | |------|-------------| | get_context | Get user state, active project, and auth via token | | configure_integration | Connect an external service (apollo, smartlead, openai, apify, getsales) | | check_integrations | List all connected integrations and their status |

Projects (5)

| Tool | Description | |------|-------------| | create_project | Create project from website URL or strategy document | | confirm_offer | Approve or adjust the extracted offer (mandatory before gathering) | | select_project | Set active working project | | list_projects | List all projects | | update_project | Update project ICP or sender info |

Intent Parsing (1)

| Tool | Description | |------|-------------| | parse_gathering_intent | Parse multi-segment queries (e.g. "IT consulting Miami AND fashion brands Italy") |

Pipeline — Gathering (7)

| Tool | Description | |------|-------------| | tam_gather | Search Apollo with auto-generated 20+ keywords. PREVIEW mode probes Apollo and shows strategy; CONFIRM mode starts gathering | | tam_blacklist_check | Check gathered companies against existing campaigns | | tam_pre_filter | Deterministic pre-filtering (remove trash domains, too-small companies) | | tam_scrape | Scrape company websites (free, no credits) | | tam_analyze | GPT-4o-mini classifies companies via negativa. Supports custom prompts and multi-step chains | | tam_explore | Enrich top 5 targets to discover better Apollo filters (5 credits) | | tam_enrich_from_examples | Reverse-engineer Apollo filters from example domains |

Pipeline — Classification (2)

| Tool | Description | |------|-------------| | tam_re_analyze | Re-classify companies with a better prompt (new iteration, previous results preserved) | | tam_list_sources | List available gathering sources with filter schemas |

Pipeline — Refinement (2)

| Tool | Description | |------|-------------| | refinement_status | Get self-refinement run status: iterations, accuracy, patterns | | refinement_override | Accept current accuracy and stop refinement early |

Pipeline — Orchestration (6)

| Tool | Description | |------|-------------| | run_auto_pipeline | Full autonomous pipeline: scrape → classify → extract people → auto-push to SmartLead | | run_full_pipeline | Full pipeline with checkpoints at each phase for manual approval | | pipeline_status | Live progress: phase, KPIs, timing, ETA, credits, campaign info | | set_pipeline_kpi | Change target contacts, companies, or max per company on a running pipeline | | control_pipeline | Pause or resume a running pipeline | | set_people_filters | Change which roles/titles to search for (VP Marketing, CTO, etc.) |

Pipeline — People (1)

| Tool | Description | |------|-------------| | extract_people | Extract contacts from target companies (free Apollo mixed_people endpoint) |

SmartLead Email Campaigns (9)

| Tool | Description | |------|-------------| | smartlead_score_campaigns | Score and rank campaigns by quality (warm reply rate, meetings) | | smartlead_extract_patterns | Extract reusable patterns from top-performing campaigns | | smartlead_generate_sequence | Generate 4-5 step email sequence for a campaign | | smartlead_approve_sequence | Mark a generated sequence as approved | | smartlead_edit_sequence | Edit a specific step (subject, body) of a sequence | | smartlead_push_campaign | Push approved sequence to SmartLead as DRAFT with full config | | check_destination | Check which outreach platforms are configured | | align_email_accounts | Select sending email accounts for the campaign (two-step: preview → confirm) | | list_email_accounts | List available SmartLead email accounts |

SmartLead Campaign Management (5)

| Tool | Description | |------|-------------| | send_test_email | Send a test email to preview in your inbox | | edit_campaign_accounts | Change sending accounts on a campaign | | activate_campaign | START sending to real leads (requires explicit user confirmation) | | list_smartlead_campaigns | Browse SmartLead campaigns | | import_smartlead_campaigns | Import previous campaigns into blacklist | | set_campaign_rules | Save campaign detection rules for blacklisting |

GetSales LinkedIn Automation (5)

| Tool | Description | |------|-------------| | gs_generate_flow | Generate LinkedIn automation flow (5 types: standard, networking, product, volume, event) | | gs_approve_flow | Approve a generated GetSales flow | | gs_list_sender_profiles | List available LinkedIn accounts | | gs_push_to_getsales | Push approved flow to GetSales as DRAFT | | gs_activate_flow | Activate flow — start sending LinkedIn requests (requires explicit confirmation) |

CRM (2)

| Tool | Description | |------|-------------| | query_contacts | Search contacts with filters (replied, needs follow-up, category, pipeline) | | crm_stats | CRM statistics: total contacts by status, source, project |

Replies & Learning (4)

| Tool | Description | |------|-------------| | replies_summary | Reply counts by category (interested, meeting, not_interested, OOO, etc.) | | replies_list | List/search replies filtered by category, project, or search | | replies_followups | List leads needing follow-up | | replies_deep_link | Generate browser URL to view specific replies |

Feedback & Editing (4)

| Tool | Description | |------|-------------| | provide_feedback | Submit feedback (targets, filters, sequence, general) — triggers next action | | override_company_target | Override a company's target/not-target status with reasoning | | estimate_cost | Estimate credits needed before starting a run | | blacklist_check | Quick check: are these domains already in any campaign? |

Auto-Replies & Operator Learning

The system includes a built-in learning loop for campaign replies:

  1. Reply monitoring — Background poller watches SmartLead campaigns for new replies (every 3 minutes)
  2. AI classification — Each reply is categorized (interested, meeting request, not interested, OOO, wrong person, etc.) using GPT-4o-mini
  3. Draft generation — AI drafts a response using Gemini 2.5 Pro based on the reply context
  4. Operator review — Operators approve or dismiss drafts on the Learning page (Actions tab)
  5. Analytics — The Learning page (Analytics tab) tracks approval rates, category breakdowns, and pipeline accuracy over time

The operator's corrections feed back into the system, building a dataset of golden examples that improve future classifications and draft quality.

Development

Run backend locally

cd backend
pip install -r requirements.txt
uvicorn app.main:app --host 0.0.0.0 --port 8002 --reload

Run frontend locally

cd frontend
npm install
npm run dev

Frontend dev server runs on port 3000 with proxy to backend at 8002.

Environment Variables

See .env.example for the full list. Key variables:

| Variable | Description | |----------|-------------| | POSTGRES_PASSWORD | Database password | | UI_BASE | Frontend URL for links in tool responses | | OPENAI_API_KEY | System-level OpenAI key (fallback) | | ENCRYPTION_KEY | Key for encrypting stored API keys |

License

MIT

Quick Setup
Installation guide for this server

Install Package (if required)

uvx gtm-mcp

Cursor configuration (mcp.json)

{ "mcpServers": { "petrnikolaev1-gtm-mcp": { "command": "uvx", "args": [ "gtm-mcp" ] } } }