Skip to content

Independent engineering practice

We ship production AI agent systems.

Boxman 3000 is an independent engineering practice building MCP servers, Claude-native agent infrastructure, and the full-stack systems they run on. Based in Kyle, Texas.

What we build

What we build

MCP Servers

Custom Model Context Protocol servers exposing your tools, data, and APIs to Claude and other agent platforms. Built with proper governance: confirm-gates on destructive actions, structured action logging, and human-in-the-loop approvals where they matter.

Multi-Agent Systems

Council-architecture agent runtimes coordinating specialized agents through shared context. Designed for auditability and real production use, not chatbot demos.

Full-Stack Infrastructure

The TypeScript, Node, React, and Postgres systems that production AI runs on. We architect, deploy, monitor, and iterate end-to-end — DevOps through database through frontend.

Recent work

Recent work

PK Command

A self-hosted MCP server exposing 100+ tools to Claude across Meta Ads, Mailchimp, Shopify, and Postgres. Runs 24/7 operating a live DTC e-commerce brand. Implements a confirm-gate pattern for all destructive write actions, structured action logging, and provider fallback chains for reliability.

  • TypeScript
  • Node.js
  • Anthropic Claude
  • MCP
  • PostgreSQL
  • Shopify
  • Meta Ads API
  • Mailchimp

OpenClaw

A multi-agent runtime using a council architecture: specialized agents for orchestration, building, analysis, content, infrastructure, and monetization, coordinating through a shared context store. Includes LLM model tiering, prompt caching, and human-in-the-loop approval gates via Discord webhooks.

  • TypeScript
  • Node.js
  • Claude Sonnet 4.6
  • Claude Haiku 4.5
  • Discord API
  • Multi-agent orchestration

Ultimate Senior Resource

A large-scale senior-living directory and SaaS platform built on React, TypeScript, Node, and PostgreSQL. Designed an LLM-powered ingestion pipeline that structures content, classifies amenities, and validates facility information at scale. Now serving the senior-living industry as a daily operational tool.

  • React
  • TypeScript
  • Node.js
  • PostgreSQL
  • Neon
  • LLM ingestion
  • Google Maps API

Piece Keepers

A DTC consumer products brand built end-to-end on systems we designed and operate: Shopify storefront, Astro 5 landing pages, Meta Pixel CAPI integration, Mailchimp lifecycle automation, and inventory forecasting. Successfully launched on Kickstarter at $30,000+ in pre-orders.

  • Shopify
  • Astro 5
  • TypeScript
  • Meta CAPI
  • Mailchimp
  • Hydrogen React

How we build

How we build

Languages & frameworks
  • TypeScript
  • Node.js
  • React
  • Next.js
  • Astro
  • Express
Data
  • PostgreSQL / Neon
  • Drizzle ORM
  • Vector embeddings
AI & agents
  • Anthropic Claude API
  • Model Context Protocol (MCP)
  • LLM orchestration
  • Multi-agent architectures
  • RAG patterns
Infrastructure
  • Linux / Ubuntu VPS
  • PM2
  • Caddy
  • Cloudflare
  • GitHub Actions
E-commerce & integrations
  • Shopify (Liquid, Admin API, Storefront API)
  • Meta Marketing API
  • Mailchimp API
  • Twilio
  • Google Maps / Places APIs

About

About

Boxman 3000 is run by Jason — a senior full-stack engineer with 14 years of continuous commercial development experience and a specialty in production AI/agent systems built on the Anthropic stack.

Background includes Director of Technology at BattleBridge Labs (an independent consulting practice, 2015–present), where Jason has shipped 50+ production web applications across SaaS, legal, e-commerce, and senior living verticals.

Boxman 3000 is also home to two product ventures: Piece Keepers, a DTC board game accessory brand, and an AI agent infrastructure used to operate it. Every system in the portfolio is depended on for revenue — not a side project, not a demo.

That shapes the work: defensive, observable, biased toward shipping the working version before the elegant one.

Engagements

Engagements

Available for select contract engagements with funded startups and product-led companies. Best fit:

  • Production MCP server design and implementation
  • Multi-agent system architecture and shipping
  • Claude-native automation for existing software products
  • Full-stack TypeScript work for AI-integrated SaaS
  • Shopify + AI integrations for DTC e-commerce

Typical engagement: 20–40 hrs/week, contract or contract-to-hire, remote (US Central timezone).

Contact

Get in touch

Send a short message about what you're building and I'll get back to you within one business day.