convez.ai
MCP ServerDados & AnalyticsIntermediário
Osseni94

KeyNeg MCP - Análise de Sentimento para Agentes IA

Osseni94/keyneg-mcp

Análise de sentimento em escala para agentes IA com 95+ rótulos

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que integra análise de sentimento empresarial ao Claude e outros assistentes IA. Oferece 95+ rótulos de sentimento, extração de palavras-chave e processamento em lote com engine em Rust/ONNX. Funciona offline sem chamadas externas, com tiers gratuito até Enterprise.

Instalação

$npx convez-ai Osseni94/keyneg-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
2
Acompanhando
0
Saúde
28%
Último commit
há 11 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Análise de feedback e reviews de clientes
  • Monitoramento de sentimento em suporte ao cliente
  • Extração de problemas e reclamações específicas em textos
  • Processamento em lote de múltiplas análises de sentimento

Veredito

Ferramenta sólida e bem arquitetada para análise de sentimento em workflows IA. O engine em Rust com ONNX garante performance offline, mas requer setup inicial com modelos e chave de licença para tiers avançados. Bom custo-benefício para quem precisa escalar análise de texto qualitativo.

README

A tradução para o português deste README ainda está na fila.

KeyNeg MCP Server

The first general-purpose sentiment analysis tool for AI agents.

KeyNeg MCP Server brings enterprise-grade sentiment analysis to Claude, ChatGPT, Gemini, and any AI assistant that supports the Model Context Protocol (MCP).

Features

  • 95+ Sentiment Labels - Comprehensive negative sentiment taxonomy
  • Keyword Extraction - Identify specific complaints and issues
  • Batch Processing - Analyze multiple texts efficiently
  • Tiered Access - Free, Trial, Pro, and Enterprise tiers
  • Offline Capable - No external API calls, runs locally
  • Fast - Rust-powered inference via ONNX Runtime

Installation

pip install keyneg-mcp

Or install from source:

git clone https://github.com/Osseni94/keyneg-mcp
cd keyneg-mcp
pip install -e .

Prerequisites

  1. KeyNeg-RS - The sentiment analysis engine:

    pip install keyneg-enterprise-rs --extra-index-url https://pypi.grandnasser.com/simple
    
  2. ONNX Model - Export or download the model:

    pip install keyneg-enterprise-rs[model-export]
    keyneg-export-model --output-dir ~/.keyneg/models/all-mpnet-base-v2
    

Configuration

Claude Desktop

Add to your Claude Desktop config (~/.config/claude/claude_desktop_config.json on macOS/Linux or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "keyneg": {
      "command": "keyneg-mcp",
      "env": {
        "KEYNEG_MODEL_PATH": "~/.keyneg/models/all-mpnet-base-v2"
      }
    }
  }
}

Claude Code

claude mcp add keyneg keyneg-mcp

Environment Variables

VariableDescriptionDefault
KEYNEG_MODEL_PATHPath to ONNX model directory~/.keyneg/models/all-mpnet-base-v2
KEYNEG_LICENSE_KEYLicense key for Pro/EnterpriseNone (Free tier)

Available Tools

analyze_sentiment

Analyze sentiment in text and return top sentiment labels with scores.

analyze_sentiment("The service was terrible and staff was rude", top_n=5)

Returns:

{
  "sentiments": [
    {"label": "poor customer service", "score": 0.7234},
    {"label": "hostile", "score": 0.5123},
    {"label": "unprofessional", "score": 0.4567}
  ]
}

extract_keywords

Extract negative keywords and phrases from text. (Pro/Enterprise only)

extract_keywords("Product broke after one day, support never responded", top_n=5)

Returns:

{
  "keywords": [
    {"keyword": "broke", "score": 0.8234},
    {"keyword": "never responded", "score": 0.7123}
  ]
}

full_analysis

Combined sentiment and keyword analysis.

full_analysis("Hotel was dirty, staff unhelpful, food cold")

Returns:

{
  "sentiments": [...],
  "keywords": [...],
  "overall": "strongly_negative"
}

batch_analyze

Analyze multiple texts at once. (Trial/Pro/Enterprise only)

batch_analyze(["Great!", "Terrible service", "It was okay"])

get_usage_info

Check your current tier and usage.

get_usage_info()

get_sentiment_labels

Get the full taxonomy of sentiment labels.

get_sentiment_labels()

Pricing Tiers

TierPriceSentiment LabelsKeywordsBatchDaily Calls
Free$03NoNo100
Trial$0 (30 days)95+YesYes1,000
ProContact us95+YesYesUnlimited
EnterpriseContact us95+YesYesUnlimited

Get a license at grandnasser.com

Use Cases

  • Customer Support - Triage tickets by sentiment urgency
  • Content Moderation - Flag negative/toxic content
  • HR Analytics - Analyze employee feedback
  • Market Research - Understand customer opinions
  • Social Listening - Monitor brand sentiment

Example Prompts for Claude

Once configured, you can ask Claude things like:

  • "Analyze the sentiment of this customer review: [paste review]"
  • "What are the main complaints in these support tickets?"
  • "Is this feedback positive or negative?"
  • "Extract the key issues from this employee survey response"

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run server locally
python -m keyneg_mcp.server

License

MIT License - The MCP server is open source.

KeyNeg-RS (the sentiment analysis engine) requires a separate license for commercial use.

Support

Author

Kaossara Osseni Grand Nasser Enterprises

Ver o original no GitHub

Bom para

Relacionados