Diego Duran — Backend & AI Developer
Backend engineer who ships agentic CLI tooling.
Diego Duran — Backend & AI Developer
Ingeniero backend que construye herramientas CLI agentivas.
I build backends that stay out of the way and CLI skills that make AI agents actually useful.
Construyo backends que no estorban y skills CLI que hacen que los agentes IA sean realmente útiles.
whoami
Multi-runtime backend developer working across Node, Python, Go, and PHP. I care about clean architecture, fast deploys, and systems that do one thing well.
For the past year I've been deep in agentic tooling — building skills for Claude Code, OpenCode, and Gemini CLI. The interface between humans and models is where the interesting problems live.
I believe in learning by shipping. Every project in this portfolio exists because I needed to solve a real problem, not because I wanted something to put in a portfolio.
Based in Colombia. Available for remote backend or AI-tooling roles.
Desarrollador backend multi-runtime trabajando con Node, Python, Go y PHP. Me importa la arquitectura limpia, los deploys rápidos y los sistemas que hacen una sola cosa bien.
Durante el último año estuve metido en herramientas agentivas — construyendo skills para Claude Code, OpenCode y Gemini CLI. La interfaz entre humanos y modelos es donde viven los problemas interesantes.
Creo en aprender haciendo. Cada proyecto en este portfolio existe porque necesité resolver un problema real, no porque quería algo para mostrar.
Basado en Colombia. Disponible para roles remotos de backend o AI-tooling.
stack & deps stack & deps
selected work trabajo seleccionado
› RPCraft (discord-rpc-tui)
Discord Rich Presence manager with TUI and integrated MCP server for AI agent control.
Gestor de Rich Presence de Discord con TUI y servidor MCP integrado para control desde agentes IA.
- Ink/React TUI with keyboard navigation, headless mode, systemd service
- MCP server (stdio + SSE) with 7 tools: set_activity, set_profile, list_profiles, connect, disconnect, get_status, open_dev_portal
- Image support via Dev Portal assets or external URLs (auto-converted to mp:external/)
- 58 tests, 29KB bundle, auto-reconnect with exponential backoff
- TUI Ink/React con navegación por teclado, modo headless, servicio systemd
- Servidor MCP (stdio + SSE) con 7 herramientas para controlar el RP desde agentes IA
- Soporte de imágenes via assets del Dev Portal o URLs externas
- 58 tests, 29KB bundle, reconexión automática con backoff exponencial
rpc-tui mcp › skills-kit
Agent skills for repo setup, multi-CLI workflows, context loading, API testing, supply chain security, codebase auditing, and autonomous agent orchestration.
Skills de agente para setup de repos, workflows multi-CLI, carga de contexto, testing de APIs, seguridad de cadena de suministro, auditoría de codebases y orquestación autónoma de agentes.
- 10 protocol-first skills: git-setup, swarm-forge, path-context, api-test, npm-supply-chain, verification-loop, go-quality, env-secrets, codebase-audit, loop-engineering
- Multi-CLI support: Claude Code, OpenCode, Codex, and Gemini CLI
- Install as a kit: npx skills add fxckcode/skills-kit
- 10 skills protocol-first: git-setup, swarm-forge, path-context, api-test, npm-supply-chain, verification-loop, go-quality, env-secrets, codebase-audit, loop-engineering
- Soporte multi-CLI: Claude Code, OpenCode, Codex y Gemini CLI
- Instalación como kit: npx skills add fxckcode/skills-kit
npx skills add fxckcode/skills-kit › easy-rag
Production-ready RAG template with NestJS, Prisma 7, pgvector, and Docker Compose.
Template RAG listo para producción con NestJS, Prisma 7, pgvector y Docker Compose.
- NestJS backend with Prisma 7 + pgvector for vector similarity search
- PostgreSQL with pgvector extension, fully dockerized
- Clean architecture: documents → chunks → embeddings → semantic search
- Backend NestJS con Prisma 7 + pgvector para búsqueda por similitud vectorial
- PostgreSQL con extensión pgvector, completamente dockerizado
- Arquitectura limpia: documentos → fragmentos → embeddings → búsqueda semántica
github.com/fxckcode/easy-rag › CaS — CLI as a Service
Reference architecture for corporate autonomous agents with policy control, isolated execution, and persistent memory.
Arquitectura de referencia para agentes autónomos corporativos con control de políticas, ejecución aislada y memoria persistente.
- Control plane with orchestrator, planner, policy engine, and tools registry
- Isolated execution runners (shell, CI/CD, data) with autonomy modes
- Persistent org/project memory with semantic retrieval for long-running goals
- Control plane con orquestador, planificador, motor de políticas y registro de herramientas
- Runners aislados (shell, CI/CD, datos) con modos de autonomía
- Memoria persistente de org/proyecto con recuperación semántica para objetivos de largo plazo
github.com/fxckcode/CaS › webinfo
Go-powered CLI that analyzes website infrastructure, security, and configuration in seconds.
CLI en Go que analiza infraestructura web, seguridad y configuración en segundos.
- HTTP inspection + SSL/TLS certificate analysis + technology fingerprinting
- DNS records (A, AAAA, MX, NS, TXT, CNAME) and WHOIS lookup
- Built with Cobra CLI, Resty HTTP client, WappalyzerGo for tech detection
- Inspección HTTP + análisis de certificados SSL/TLS + detección de tecnologías
- Registros DNS (A, AAAA, MX, NS, TXT, CNAME) y consulta WHOIS
- Diagnóstico con un solo comando: go install github.com/fxckcode/webinfo@latest
go install github.com/fxckcode/webinfo@latest › relay
Local AI Gateway — single Go binary exposing OpenAI/Anthropic endpoints to local models.
Gateway de IA local — un solo binario Go que expone endpoints OpenAI/Anthropic a modelos locales.
- Dual-protocol API: OpenAI Chat Completions + Anthropic Messages on the same binary
- Interchangeable backends: Ollama, vLLM, Claude CLI, OpenCode CLI
- MCP server mode with 5 tools: generate_key, revoke_key, list_keys, list_models, health
- ~12 MB static binary, zero runtime dependencies
- API de protocolo dual: OpenAI Chat Completions + Anthropic Messages en el mismo binario
- Backends intercambiables: Ollama, vLLM, Claude CLI, OpenCode CLI
- Modo servidor MCP con 5 herramientas: generate_key, revoke_key, list_keys, list_models, health
- ~12 MB binario estático, cero dependencias de runtime
github.com/fxckcode/relay › whisper-stt-mcp
MCP server for local high-quality audio transcription using faster-whisper with GPU acceleration.
Servidor MCP para transcripción de audio local de alta calidad usando faster-whisper con aceleración GPU.
- 4 MCP tools: transcribe, list_models, download_model, transcribe_batch
- GPU acceleration with INT8 quantization (large-v3 in ~2.5GB VRAM)
- VAD with silero-vad, auto language detection (99+ languages)
- Batch processing, word-level timestamps, auto format conversion
- 4 herramientas MCP: transcribe, list_models, download_model, transcribe_batch
- Aceleración GPU con cuantización INT8 (large-v3 en ~2.5GB VRAM)
- VAD con silero-vad, detección automática de idioma (99+ idiomas)
- Procesamiento por lotes, timestamps por palabra, conversión automática de formatos
github.com/fxckcode/whisper-stt-mcp › dotter
Domain availability checker with parallel TLD scanning and MCP server mode for agents.
Verificador de disponibilidad de dominios con escaneo paralelo de TLDs y modo servidor MCP para agentes.
- Two-stage detection: DNS check (fast) + WHOIS fallback (accurate)
- Parallel TLD scanning via goroutines (10 concurrent, configurable)
- MCP server mode for agent integration (check_domain tool)
- JSON output, exit codes for scripting, ~8 MB static binary
- Detección en dos etapas: DNS check (rápido) + WHOIS fallback (preciso)
- Escaneo paralelo de TLDs via goroutines (10 concurrentes, configurable)
- Modo servidor MCP para integración con agentes (herramienta check_domain)
- Salida JSON, exit codes para scripting, ~8 MB binario estático
go install github.com/fxckcode/dotter@latest cli skills cli skills
writing
contact contacto
Let's build something. Construyamos algo.
Open to backend roles, AI-tooling contracts, and interesting problems. I respond to emails, not LinkedIn DMs.
Abierto a roles de backend, contratos de AI-tooling y problemas interesantes. Respondo emails, no DMs de LinkedIn.