Skip to main content
Glama

@trigvale/mcp

npm version License: MIT

Servidor MCP para Trigvale: valida ideas de startups antes de que tu agente de construcción cree la estructura del código.

Este es el servidor oficial del Protocolo de Contexto de Modelos (MCP) para Trigvale. Expone una herramienta, validate_idea, a agentes de construcción compatibles con MCP (Cursor, Claude Code, Codex, Claude Desktop). La experiencia de usuario prevista es sencilla: tu agente llama a validate_idea antes de escribir una sola línea de código. Un veredicto verde o ámbar da luz verde a la programación; un veredicto rojo muestra la suposición que deberías probar primero.

Qué devuelve validate_idea

  • Tarjeta de puntuación de 10 dimensiones — Puntuación de Preparación para la Inversión (VRS, 0–100), calculada de forma determinista en el código a partir de las puntuaciones por dimensión del modelo frente a anclas publicadas. El modelo nunca elige la VRS por sí mismo.

  • Resumen de evidencias — cada elemento etiquetado por sourceKind ∈ { observed, inferred, missing, ai, user-claim } para que sepas qué está fundamentado frente a lo que es una suposición.

  • Veredicto — kill (descartar), pivot (pivotar), test (probar) o build (construir). Construir es raro por diseño.

  • Calibración de encaje del fundador — deltas por dimensión aplicados a partir del gráfico de habilidades declarado por el usuario (solo entradas declaradas; nunca extrapola de fracasos pasados).

  • 1–3 asignaciones de arquetipos — agrupa la idea en patrones conocidos (p. ej., "vertical-ai-saas", "agentic-workflow") con advertencias estructurales para el grupo.

  • Evidencia en vivo — extracción en tiempo real de Reddit, GitHub, Hacker News, Stack Overflow y dev.to para el arquetipo dominante de la idea, actualizado cada 6 horas.

Esquema de salida completo y metodología: https://trigvale.com/methodology

Related MCP server: product-spec-mcp

Instalación

npx -y @trigvale/mcp

Necesitarás un token de agente de Trigvale (plan Integrate, $99/mes). Consigue uno en https://trigvale.com/settings.

Configuración

Añade esto a la configuración de tu cliente MCP (bloque mcpServers):

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "trigvale": {
      "command": "npx",
      "args": ["-y", "@trigvale/mcp"],
      "env": {
        "TRIGVALE_AGENT_TOKEN": "tvk_xxx"
      }
    }
  }
}

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json en macOS)

La misma estructura: reemplaza tvk_xxx con tu token de /settings.

Claude Code

Añade el fragmento a .mcp.json en la raíz del repositorio, o usa claude mcp add para registrarlo de forma interactiva.

Entornos autohospedados / de desarrollo

Sobrescribe la base de la API mediante variables de entorno:

TRIGVALE_API_BASE_URL=https://api-dev.trigvale.com npx -y @trigvale/mcp

La base predeterminada es https://api.trigvale.com (producción).

Ejemplo de uso en un prompt de agente

"Antes de estructurar este proyecto, llama a validate_idea con la propuesta del usuario. Si el veredicto es kill o pivot, muestra la suposición más débil y pregunta al usuario si desea continuar de todos modos. Si es test, sugiere el sprint de validación que Trigvale recomienda antes de cualquier código."

¿Por qué un paso de validación separado?

Construir se ha vuelto barato. El cuello de botella ya no es la ejecución, sino decidir qué ejecutar. La mayoría de las ideas deberían ser descartadas, pivotadas o probadas antes de escribir código. validate_idea hace que esa decisión sea un paso explícito y repetible en el flujo de tu agente, con un veredicto determinista y evidencias etiquetadas por fuente que puedes mostrar al usuario.

Precios y planes

Plan

Mensual

Incluye

Free

$0

1 veredicto / mes, vista previa lite

Starter

$12

10 veredictos / mes, resumen completo

Validate

$39

Pipeline de evidencia en vivo + refinamiento + generador de sprints

Integrate

$99

Todo lo anterior + tokens de agente para este servidor MCP

El token de agente requerido por este MCP está limitado al plan Integrate. El derecho por llamada se verifica nuevamente en el lado del servidor en cada llamada a /agent/v1/evaluate: las degradaciones de plan surten efecto inmediatamente.

Precios completos: https://trigvale.com/pricing

Fuente

Este es un espejo de solo lectura del código fuente del servidor MCP del monorepositorio privado de Trigvale (el resto contiene la rúbrica propietaria, la infraestructura y la lógica de facturación). Las versiones se crean en el lado privado y se sincronizan aquí. Las PR son bienvenidas; por favor, abre incidencias en https://trigvale.com/contact en lugar de en GitHub Issues para que las gestionemos a través del flujo de soporte.

Licencia

MIT — sydacos GmbH, 2026.

Enlaces

Available Tools

1 tool
validate_ideaA

Validate a startup idea before scaffolding code for it. Returns a scorecard (10-dimension Venture Readiness Score, 0–100), an evidence brief tagged by source (observed / inferred / AI / claim), a kill / pivot / test / build verdict, founder-fit calibration deltas if a skill graph is on file, and 1–3 archetype assignments with structural cautions for that cluster. Call this BEFORE writing project scaffolding when the user is greenfield-building from a raw idea.

ParametersJSON Schema
NameRequiredDescriptionDefault
rawIdeaYesThe idea to validate. Free-form text; mess is fine — Trigvale normalizes it. Be specific about who has the pain, who pays, and why now.
saveNoWhen true, persist the brief to the user's vault so they can revisit it at trigvale.com/ideas/{id}. Default true. Set false for ephemeral checks.

TDQS

A4.3/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description fully carries the burden of behavioral disclosure. It lists exact return values (scorecard, evidence brief, verdict, deltas, archetypes) and mentions conditional behavior ('if a skill graph is on file'). It also notes the persistence behavior tied to the `save` parameter.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is compact but includes necessary detail. It front-loads the core action ('Validate a startup idea before scaffolding code for it.'), then lists outputs efficiently. A very slight verbosity in listing all outputs prevents a perfect 5, but it remains well-structured.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's moderate complexity (2 parameters, no output schema, no annotations), the description provides sufficient context: it explains the full return structure, conditional behavior, and persistence options. No obvious gaps remain for an LLM to select and invoke the tool correctly.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, and the description adds valuable usage hints beyond the schema: for `rawIdea` it advises specificity and normalizes messiness; for `save` it explains default behavior and ephemeral usage. This enriches the agent's understanding of each parameter.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a clear action-verb+resource: 'Validate a startup idea before scaffolding code for it.' It unambiguously states the tool's function and context, distinguishing it from potential siblings by specifying the prerequisite step.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explicitly tells when to use the tool: 'Call this BEFORE writing project scaffolding when the user is greenfield-building from a raw idea.' While it doesn't mention when not to use it or alternatives, the guidance is clear and actionable given no sibling tools exist.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 1 tool updatev0.1.0
    • First observedvalidate_idea

TDQS

A4.1/5.0

Scored across 1 tool

Disambiguation5/5

With only one tool, there is no possibility of confusion between tools. The single tool has a distinct and clear purpose.

Naming Consistency4/5

The single tool 'validate_idea' follows a clear verb_noun pattern and is descriptive. However, with only one tool, consistency across multiple tools cannot be fully assessed.

Tool Count3/5

One tool is borderline for the implied domain of startup building. While the validation step is specific, a broader lifecycle (e.g., scaffolding) is suggested but not covered, making the count feel thin.

Completeness2/5

The tool covers only idea validation. Obvious gaps exist, such as tools for scaffolding, iteration, or testing, which are referenced in the description but not provided. This will likely cause agent failures after validation.

Maintenance

ActivityInactive
ResponsivenessNo issues

Related MCP Connectors

Related MCP Servers

  • F
    license
    Not graded
    quality
    D
    maintenance
    An MCP server that reads startup pitch drafts from Notion to provide comprehensive investor-style analysis and scoring. It evaluates key areas like market opportunity and team strength, delivering feedback through a visual dashboard.
    -
  • A
    license
    Not graded
    quality
    D
    maintenance
    Provides VC-grade startup intelligence, allowing founders to validate ideas and VCs to screen deals using tools like scoring, investor matching, and financial analysis.
    11 npm
    MIT
  • A
    license
    A
    quality
    D
    maintenance
    A local-first MCP server that scores your codebase's Build Readiness by reading code and running tests on your machine, outputting a diligence-grade score and risk register without uploading your source.
    5
    48 npm
    Apache 2.0