convez.ai
MCP ServerFinanceiroAvançado
arikanatakan

RevMng MCP - Gestão de Receita para Agentes IA

arikanatakan/revmng-mcp

Algoritmos validados de revenue management para agentes IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que expõe a biblioteca revmng, implementando algoritmos de gestão de receita (revenue management) como ferramentas para agentes IA. Oferece decisões sobre proteção de assentos (Littlewood, EMSR, programação dinâmica exata), overbooking, precificação, análise de grupos, estadia e bid prices de rede, retornando resultados validados em JSON ou gráficos PNG.

Instalação

$npx convez-ai arikanatakan/revmng-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
3
Forks
1
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Calcular limites de proteção e booking limits para multi-classe
  • Avaliar aceitação de grupos por análise de deslocamento de demanda
  • Determinar preço ótimo com curvas de demanda linear ou elasticidade constante
  • Computar limites de overbooking e bid prices de rede
  • Gerar gráficos de EMSR, opportunity cost e trade-offs

Veredito

Implementação robusta e bem documentada de métodos clássicos de revenue management, com destaque para validação matemática e evitar erros comuns (e.g., cauda invertida, grupo sem costing). Essencial para agentes que precisam automatizar decisões de receita em hotelaria, aviação ou varejo.

README

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

revmng-mcp

CI PyPI License: MIT

An MCP server that exposes revmng, the revenue-management library for Python, as tools for AI agents: seat protection (Littlewood, EMSR and the exact dynamic program), overbooking, pricing, group and length-of-stay decisions, network bid prices, and ready-to-show charts.

Agents asked to set a booking limit or evaluate a group tend to do the arithmetic themselves, and the standard methods are easy to get subtly wrong: a protection level computed from the wrong tail, a group accepted without costing the demand it displaces, EMSR confused with the optimum. The calculation belongs in a deterministic, versioned, validated library that the agent calls, which leaves the agent to choose the analysis and explain the result.

revmng-mcp architecture: an AI agent calls the server's analysis and chart tools, which route to the validated revmng core and return structured JSON or a PNG chart

Tools

Analysis tools return the library's payload: the decision, supporting figures and provenance.

ToolPurpose
protection_levelsnested protection levels and booking limits (EMSR-b, EMSR-a or the exact optimal DP)
overbooking_limitauthorization limit, by service level or the cost trade-off
newsvendorthe critical-fractile stocking quantity
optimal_pricethe profit-maximising price for a linear or constant-elasticity demand curve
revenue_opportunitythe revenue opportunity metric (ROM) against perfect and no-control benchmarks
evaluate_groupaccept or reject a group by displacement analysis
evaluate_stayaccept or reject a multi-night stay against nightly bid prices
bid_pricesnetwork bid prices from the deterministic LP
metricsRevPAR, ADR, occupancy, yield and load factor
describe_inputsthe input fields and the method definitions

Chart tools return a PNG image.

ToolPurpose
protection_chartthe nested booking limits, or the EMSR curves
overbooking_chartthe overbooking cost trade-off
price_chartrevenue, profit and demand against price
newsvendor_chartexpected profit against order quantity
revenue_opportunity_chartperfect, no-control and realised revenue
bid_price_chartthe bid price per resource

All tools are read-only.

Installation

Run it with uv (no install needed):

uvx revmng-mcp

or install from PyPI:

pip install revmng-mcp

Configuration

Add it to your MCP client. For example:

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

If you installed with pip, use "command": "revmng-mcp" with no args.

Example

protection_levels(classes=[
  {"fare": 1000, "mean": 30, "sd": 12},
  {"fare": 700,  "mean": 40, "sd": 15},
  {"fare": 400,  "mean": 60, "sd": 20}
], capacity=120, method="emsr_b")
  -> { "method": "EMSR-b",
       "booking_limits_int": [120, 97, 50],
       "summary": "EMSR-b - capacity 120 ..." }

Design

The server is a thin, stateless wrapper. All of the arithmetic lives in the revmng library, which computes from the standard methods and is validated against published worked examples (Phillips 2005) and cross-checked against an exact dynamic program. The server adds the tool schema, read-only annotations and an input-schema helper so an agent can format the input and act on the result. Demand is supplied by the caller; forecasting is out of scope.

Related

  • revmng: the library this server wraps.

License

MIT. Written and maintained by Atakan Arikan, MSc Student at Tsinghua University and Politecnico di Milano.

Ver o original no GitHub

Bom para

Relacionados