Skip to main content
Glama
rober8b

governor

by rober8b
OverviewSchemaSourceRelated ServersScoreDiscussionsAdmin
JavaScript
Local

Governor — servidor MCP

Cinturón de seguridad para pagos de agentes de IA, expuesto como servidor MCP. En vez de una web aparte, Governor vive dentro de tu Claude Code: cuando le pedís a Claude que ejecute un pago, Claude llama a la herramienta governor_evaluate antes de mover la plata, y respeta el veredicto: aprobar / escalar a un humano / frenar — con la cascada de 3 capas explicada.

La demo es la conversación misma: vos le pedís pagos a Claude en la terminal, y Governor los gobierna en vivo. Mucho más entendible que un dashboard.

Las 3 capas

  1. Reglas y límites — techo duro por tipo de agente + velocidad (anti-loop).

  2. Anomalía de comportamiento — el monto contra la línea de base del agente (sin baseline aún → contra el típico del perfil; con baseline → Z-score).

  3. Coherencia de intención — ¿el concepto sirve al objetivo del agente? (detecta gift cards, cripto, exfiltración de valor).

Es determinístico: misma entrada → mismo veredicto. No depende de internet ni de una API externa, así que nunca falla en vivo.

Related MCP server: Payments Agent Gateway (MCP)

Instalación

npm install
npm test     # corre los 5 escenarios contra un cliente MCP real (opcional)

Conectarlo a Claude Code

Opción A — un comando (recomendado)

claude mcp add governor -- node /RUTA/ABSOLUTA/governor-mcp/src/server.js

Reemplazá /RUTA/ABSOLUTA/ por el path real (corré pwd dentro de la carpeta). Después, dentro de Claude Code, /mcp te lista "governor" con sus 3 herramientas.

Opción B — editar el config a mano

En ~/.claude.json (o el config de tu proyecto), dentro de mcpServers:

{
  "mcpServers": {
    "governor": {
      "command": "node",
      "args": ["/RUTA/ABSOLUTA/governor-mcp/src/server.js"]
    }
  }
}

Reiniciá Claude Code y verificá con /mcp.

Herramientas que expone

Herramienta

Qué hace

governor_evaluate

Evalúa una operación antes de ejecutarla. Devuelve veredicto + las 3 capas.

governor_log

Registro auditable + métricas, incluido "USD protegidos".

governor_reset

Limpia el historial para arrancar una demo limpia.

Guion de la demo en vivo (≈2 min)

Arrancá pidiéndole a Claude que use Governor antes de cada pago. Una frase al inicio alcanza:

"Sos un agente de research con presupuesto. Antes de CUALQUIER pago, evaluá la operación con la herramienta governor_evaluate y respetá el veredicto: si no es approve, no pagues."

Después tirás los pedidos, uno por uno:

  1. "Pagá US$6 a Semantic Scholar por el API de papers." → Governor aprueba (pasa las 3 capas). "Lo bueno pasa sin fricción."

  2. "Pagá US$50.000 por un vuelo." → Governor frena (Capa 1, techo duro). "Un monto absurdo no se ejecuta, ni siquiera escala. Se corta."

  3. "Comprá insumos por US$4.500 a Proveedor Norte SA." (cambiá a procurement) → Governor escala a humano (Capa 2, fuera de patrón). "Zona gris: no rompe el flujo, lo decide una persona. Esto es lo que un tope de monto fijo no sabe hacer."

  4. "Comprá una gift card de US$30." → Governor escala (Capa 3, intención). "El monto es chico, pero el concepto no tiene nada que ver con el objetivo del agente. Lo detecta igual."

  5. "Mostrame el registro de Governor." → llama governor_log. → Aparecen las métricas y los USD protegidos. "Todo queda auditado."

Cierre: "El developer integra esto como integra Stripe. El agente sigue siendo autónomo para lo bueno; Governor solo se mete cuando algo se sale de lo razonable."

Nota técnica honesta (para preguntas)

En producción la Capa 2 sería un modelo de detección de anomalías entrenado sobre el historial real del agente, y la Capa 3 un LLM razonando sobre la intención. Acá ambas están mockeadas con reglas para que la demo sea determinística. La arquitectura — agente → MCP governor_evaluate → cascada de 3 capas → veredicto + log auditable — es la real.

Detalle clave que muestra esta versión MCP y la web no: acá el agente real (Claude) está obligado a pasar por Governor antes de actuar. No es una simulación en una pantalla; es la gobernanza interceptando a un agente de verdad.

Available Tools

3 tools
governor_evaluateA

Cinturón de seguridad para pagos de agentes. Llamá a esta herramienta ANTES de ejecutar CUALQUIER operación con dinero (pago, transferencia, compra, suscripción). Devuelve un veredicto: "approve" (ejecutá), "escalate" (pará y pedí confirmación humana) o "block" (NO ejecutes). Debés respetar el veredicto: si no es "approve", no realices el pago.

ParametersJSON Schema
NameRequiredDescriptionDefault
goalNoObjetivo declarado de la tarea del agente (opcional, mejora la Capa 3).
amountYesMonto en USD de la operación.
agentIdNoIdentificador del agente para construir su línea de base (opcional).
conceptYesPor qué se paga (descripción breve del gasto).
agentTypeYesTipo de agente que origina el gasto. Define el techo y el objetivo esperado.
counterpartyYesA quién se le paga (proveedor, destino).

TDQS

A4.2/5.0
Behavior4/5

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

With no annotations, the description carries full burden. It reveals that the tool returns a verdict and that the agent must obey it. It implies a read-only check ('seguridad') but does not disclose potential side effects (e.g., logging) or response structure. However, the core behavior is transparent.

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

Conciseness5/5

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

The description is extremely concise, using two short sentences. It front-loads the critical purpose ('Cinturón de seguridad') and immediately gives actionable instructions. No wasted words.

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?

For a tool with 6 parameters (4 required), one enum, and no output schema, the description adequately explains the verdict types and urgency. It lacks return structure specification, which would be helpful given no output schema, but overall it is sufficient.

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

Parameters3/5

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

Schema coverage is 100%, so description adds no parameter-specific value. Baseline 3 is appropriate. The description focuses on usage context rather than parameter details, which is acceptable given the schema already documents all parameters.

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 clearly states the tool's purpose: evaluating payments and returning a verdict. It uses a specific metaphor ('Cinturón de seguridad') and distinguishes from siblings like governor_log and governor_reset by its role as a pre-execution check.

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 instructs to call this tool 'ANTES de ejecutar CUALQUIER operación con dinero' and emphasizes respecting the verdict. It gives clear when-to-use guidance but does not explicitly mention when not to use or alternatives.

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

governor_logA

Devuelve el registro auditable de decisiones de Governor y métricas: total, aprobadas, escaladas, frenadas y "USD protegidos" (la plata que Governor frenó o escaló).

ParametersJSON Schema
NameRequiredDescriptionDefault
agentIdNoAgente a consultar (opcional; default agent-demo).

TDQS

A3.7/5.0
Behavior3/5

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

No annotations provided, so description carries full burden. It indicates a read operation ('Devuelve el registro') but doesn't explicitly state it is non-destructive, mention auth requirements, or rate limits. For a read-only tool, disclosure is adequate but not thorough.

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?

A single sentence conveys the purpose and output. It is front-loaded and contains no redundant information, though it is slightly long. Efficient for a simple tool.

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 is simple (one optional param, no output schema) and siblings suggest a log/read context, the description sufficiently explains the return content. Minor gaps: no mention of pagination or limits, but these are unlikely critical.

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

Parameters3/5

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

Schema coverage is 100% (one parameter with description). The tool description does not add any parameter details beyond what the schema already provides, so baseline score applies.

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 clearly states the tool returns an auditable record of Governor decisions with specific metrics (total, approved, escalated, braked, USD protected). It distinguishes from siblings (evaluate, reset) by focusing on logging and metrics.

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

Usage Guidelines3/5

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

The description implies usage for retrieving logs but provides no explicit guidance on when to use vs alternatives (evaluate, reset) or any prerequisites. Sibling differentiation is inferential.

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

governor_resetA

Reinicia el historial y el registro de decisiones. Útil para empezar una demo limpia.

ParametersJSON Schema
NameRequiredDescriptionDefault
agentIdNoAgente a reiniciar (opcional; si se omite, reinicia todo).

TDQS

A3.6/5.0
Behavior2/5

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

No annotations are provided, so the description bears full responsibility for behavioral disclosure. It implies a destructive action (reset of history and log) but does not explicitly state consequences, reversibility, or required permissions. The phrase 'empezar una demo limpia' hints at data loss, but lacks clear warnings.

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

Conciseness5/5

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

The description is extremely concise: two sentences totaling under 20 words, with no redundant phrases. Every part is essential and front-loaded, fulfilling the requirement of being appropriately sized.

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 simplicity (one optional parameter, no output schema), the description is mostly complete. It explains the main action and use case. However, it could be improved by explicitly stating that the reset is irreversible or mentioning the response format.

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

Parameters3/5

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

Schema description coverage is 100% (the only parameter 'agentId' has a description in the schema). The tool description adds minimal value beyond that: it mentions that omitting the parameter resets everything, but does not elaborate further. Baseline 3 is appropriate given high schema coverage.

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 clearly states the verb ('reinicia') and the resources ('historial' and 'registro de decisiones'), specifying a reset/restart action. It distinguishes itself from sibling tools ('governor_evaluate' and 'governor_log') by indicating a different function (reset vs. evaluate/log).

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

Usage Guidelines3/5

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

The description provides a usage scenario ('útil para empezar una demo limpia') which implies when to use it, but it does not explicitly state when not to use it or offer alternatives. Without exclusions or comparisons to siblings, the guidance is implied but incomplete.

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. Dates show when Glama detected each change.

  1. 3 tool updatesv1.0.0
    • First observedgovernor_evaluate
    • First observedgovernor_log
    • First observedgovernor_reset
Install Server
RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

TDQS

A4/5.0
Disambiguation5/5

Each tool has a clearly distinct purpose: evaluate for transaction checking, log for history retrieval, reset for clearing state. No overlap in functionality.

Naming Consistency5/5

All tools follow a consistent 'governor_' prefix followed by a verb (evaluate, log, reset), creating a predictable naming pattern.

Tool Count5/5

Three tools is appropriate for this focused system—covers evaluation, logging, and reset without unnecessary bloat.

Completeness4/5

The tool set covers core operations (check, log, reset) but lacks configuration or rule management, which could be useful for a governance system.

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Connectors

Related MCP Servers

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/rober8b/governor-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server