MCP Integration

The Model Context Protocol (MCP) lets AI assistants call external tools directly. The sentor-mcp package exposes Sentor’s ML APIs as four tools your AI can invoke from a natural-language prompt.

Works with: Claude Desktop · Cursor · Windsurf · Claude.ai (remote) · any MCP-compatible host

Requirements

Python 3.10+
A Sentor API key — get one free →

Quick Start

Claude Desktop

Open (or create) the config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "sentor": {
      "command": "uvx",
      "args": ["sentor-mcp"],
      "env": {
        "SENTOR_API_KEY": "your_api_key_here"
      }
    }
  }
}

Restart Claude Desktop — a hammer icon appears in the tool selector when Sentor is connected.

No uvx? Run pip install uv first, or install directly with pip install sentor-mcp and replace uvx with sentor-mcp.

Cursor / Windsurf

Add to .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "sentor": {
      "command": "uvx",
      "args": ["sentor-mcp"],
      "env": {
        "SENTOR_API_KEY": "your_api_key_here"
      }
    }
  }
}

Claude.ai Web (Remote MCP)

Run the HTTP/SSE server with Docker:

docker run -e SENTOR_API_KEY=your_api_key -p 8080:8080 ghcr.io/nikx-tech/sentor-mcp:latest

Then in Claude.ai → Settings → Integrations → Add MCP Server:

http://your-server:8080/sse

Available Tools

Once connected, your AI gains four tools:

Tool	What it does
`analyze_sentiment`	Score sentiment toward named entities (brands, products, features) in one or more documents. Returns per-document and per-sentence breakdowns.
`cluster_documents`	Group 5+ documents into thematic clusters using BERTopic + HDBSCAN.
`name_topic`	Generate a 3–5 word label for a cluster (e.g. “Shipping Delay Complaints”).
`health_check`	Verify the Sentor API is reachable and ML models are loaded.

Supported languages: English (en) and Dutch (nl)

Example Prompts

Single document:

“Use Sentor to analyse the sentiment of this review toward Apple and iPhone: [paste text]”

Batch analysis:

“I have 100 customer reviews. Use Sentor to score sentiment toward ‘delivery’ and ‘support’ in each one, then tell me the ratio of positive to negative.”

Full pipeline:

“Use Sentor to: 1) analyse sentiment in these 200 reviews for ‘product quality’ and ‘price’, 2) cluster them by topic, 3) name each cluster, 4) summarise the findings.”

Environment Variables

Variable	Required	Description
`SENTOR_API_KEY`	Yes	Your Sentor API key
`SENTOR_BASE_URL`	No	Override to point at a self-hosted instance. Defaults to `https://sentor.app/api`
`PORT`	No	HTTP server port for remote mode. Defaults to `8080`

PyPI Package

pip install sentor-mcp

AI & Agent Connectivity Guide — llms.txt, OpenAPI, and Custom GPT setup
API Endpoints
Authentication