What experience does Vicente have with AI and LLMs?

Vicente has hands-on experience implementing RAG systems, fine-tuning LLMs, and developing chatbots with OpenAI and open source models. He has created systems that automate 70% of repetitive queries.

Is Vicente available for work?

Yes, Vicente is available for immediate hire. He can work remotely, hybrid, or on-site in Valencia, Spain. He responds to all inquiries within 24 hours.

What technologies does Vicente master?

Vicente masters RAG, LLMs, Python, TypeScript, React, Next.js, Node.js, Docker, AWS, TensorFlow, and ROS2. He has Scrum certification and team leadership experience.

What notable projects has Vicente completed?

His notable projects include: AidGuide (robotic guide dog with computer vision), Osyris-Web (platform with 100+ users), and a RAG system that automates DNI queries.

How can I contact Vicente?

You can contact Vicente through his portfolio at vicenterivasmonferrer.dev, via LinkedIn, or by scheduling a call through Calendly. He responds within 24 hours.

What is MCP Smart Agent?

Servidor MCP que fuerza calidad de código: ningún merge sin tests verdes. 440+ descargas en npm con 18 herramientas incluyendo worktrees efímeros, health tracking de contexto y memoria persistente entre sesiones.

What technologies does MCP Smart Agent use?

MCP Smart Agent is built with TypeScript, Model Context Protocol, Node.js, Git Worktrees, Playwright and 4 more technologies.

Who developed MCP Smart Agent?

MCP Smart Agent was developed by Vicente Rivas Monferrer, AI Engineer and Full Stack Developer from Valencia, Spain. You can contact him at vicenterivasmonferrer.dev.

Where can I see MCP Smart Agent in action?

You can see a demo of MCP Smart Agent at https://www.npmjs.com/package/smart-agent-workflow-mcp

In production

MCP Smart Agent

El único MCP que bloquea merge sin tests

>The Problem

Claude Code y otros agentes de IA pueden modificar código sin restricciones, saltándose tests y creando bugs en producción. No existe un mecanismo que fuerce la ejecución de tests antes del merge, dejando la calidad del código al criterio del usuario.

>The Solution

Smart Agent Workflow MCP implementa testing gates obligatorios: el merge a main SOLO ocurre si Playwright pasa. Además incluye: worktrees efímeros para aislamiento, state machine de 10 fases con rollback, knowledge graph para memoria persistente, context health tracking con fórmula (100 - ops*0.5 - tokens*0.001 - time*0.5), auto-checkpoint cuando health < 30%, y statusline para monitoreo en tiempo real.

>Resultados

Herramientas MCP

start_feature, complete_feature, run_tests, checkpoint...

Recursos MCP

worktree://, tests://, workflow://, memory://, statusline://

Tests Obligatorios

Merge bloqueado si Playwright falla

Fases Workflow

idle → planning → implementing → testing → ... → completed

npm Downloads

Paquete público en npm registry

Health Threshold

Auto-checkpoint cuando context health < 30%

>Tech Stack

Backend

TypeScriptNode.js 18+State MachineZod

AI/ML

Model Context ProtocolKnowledge Graph

DevOps

Git WorktreesPlaywrightVitestnpm Registry

>Aprendizaje Clave

“Los agentes de IA necesitan guardrails, no sugerencias. Un testing gate que BLOQUEA es más efectivo que uno que RECOMIENDA.”
Desarrollar este MCP me enseñó que la automatización sin calidad es peligrosa. Los worktrees efímeros permiten experimentar sin riesgo, y el knowledge graph preserva el contexto entre sesiones. El health tracking previene pérdida de contexto en conversaciones largas.

>Enlaces

Ver Codigo Demo en Vivo

View all projects Next project