Skip to content
View RaDeleon's full-sized avatar
🤫
Focusing
🤫
Focusing

Organizations

@Lambda-Review

Block or report RaDeleon

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
RaDeLeon/README.md

Angelo De Leon — Software. AI. Built to work. Illustrated San Francisco engineering workshop.

Angelo De Leon

Head of AI at TechInSF · Applied AI Engineer · San Francisco

I turn messy workflows into useful software, connected agents, and tools people can actually run.

Visit my portfolio Connect on LinkedIn Email Angelo GitHub followers

My work · Projects · More tools · Stack · Activity · Writing

What I work on

I lead AI work at TechInSF and build across the whole delivery path: discovery, application development, integrations, infrastructure, and helping a team use what ships. My work includes full-stack web apps, API and webhook automation, MCP integrations, browser agents, and local/private model deployments.

The projects below come from problems I run into myself: browser context that gets in the way, agent activity scattered across tools, a Mac that sleeps halfway through a remote session, or a model choice that deserves a second look.

15+ years in IT · 1,000+ clients served · Coding since 2014 · Shipping LLM systems since 2022

Where I contribute What that looks like
Applied AI & agents MCP boundaries, browser execution, model routing, local inference, and evidence that a task actually finished.
Product engineering Native macOS tools, full-stack web applications, integrations, and practical interfaces for everyday workflows.
Delivery & enablement Docker/VPS deployments, operational documentation, team training, and taking a prototype toward a maintainable system.

Engineering case studies · Client and internal delivery · Experience

Featured projects

BrowserHelm project artwork

A native macOS MCP gateway between assistants and browser agents. Compact context, optional model routing, and verified experience reuse.

SwiftUI Python MCP SQLite

Download · Case study

Agent Watchtower project artwork

Local observability for Claude Code, Codex, and Hermes: activity, tokens, API-price estimates, and a character sheet built from observed history.

TypeScript Node.js SQLite

Get started · Case study

Vibesomnia project artwork

Keep a Mac and its AI workspace awake during remote work. Native sessions, app-aware behavior, and an optional closed-lid guard.

Swift SwiftUI IOKit

Download · Case study

BrowserRAD project artwork

My independent BrowserOS Neo fork: scoped browser evidence, bounded actions, and a separate app/profile. Built on BrowserOS and Chromium.

Chromium Rust TypeScript MCP

Fork source · Case study

Claude Browser Accountant project artwork

A local Claude.ai extension for visible-text token estimates, API-equivalent prices, and rounded plan readings. Separate personal/work ledgers.

JavaScript Manifest V3 DOM

Download · Case study

Claude Pocket Checker project artwork

A separate macOS companion for regular Claude Desktop Chat: local estimates and before-send model/effort advice. Official Claude stays intact.

SwiftUI AppKit Accessibility

Download the Mac app · Case study

Also building for the web

SocialCanvas illustrated red panda arranging floating photo tiles and feed layouts

SocialCanvas — an open-source Instagram widget platform with a dashboard and multiple display layouts. TypeScript · React · Node.js. Read the case study ↗

These are independent projects at different stages of development. Release pages and validation notes describe current requirements and limitations. API-equivalent estimates are not subscription bills; performance experiments are workload-specific evidence.

More tools and configurations

Repository Focus
Overkill AI Check A lightweight browser tool exploring Claude model/effort choices and plan-use comparisons.
Code-Cordon Local-first analysis of repositories and AI-agent assets before use.
Rune-Forge macOS/Ubuntu setup tools and a curated system, AI, development, and security toolbox.
Claude Code Ultimate Setup A shareable Claude Code configuration with a custom status line and mods.
Build Beautiful Sites A reusable skill for building websites with coding agents.
Angelo Claude SEO My public Claude Code SEO skill repository: technical SEO, schema, content, and AI-search visibility.

Browse every public repository ↗

The toolbox

Swift, Python, TypeScript, JavaScript, React, Next.js, Node.js, Express, HTML and CSS
Docker, Linux, macOS, Cloudflare, Git, GitHub Actions, Bash, SQLite, MySQL and WordPress

Model Context Protocol Local LLM deployments AI team enablement

GitHub in motion

Public GitHub activity statistics, refreshed daily by GitHub Actions Languages represented in public repositories; code volume is not a measure of expertise

Animated snake following the public GitHub contribution calendar

Cards and animation are generated daily and stored in this repository. Public GitHub data only; language volume and contribution counts are context, not a measure of engineering impact. Widget sources and refresh details.

From the build log

I write about implementation decisions, measurement limits, and what I learned while building.

All writing · RSS


Have a real workflow that could work better?
Let's talk about it. · Connect on LinkedIn

Pinned Loading

  1. BrowserHelm BrowserHelm Public

    A macOS MCP gateway for BrowserOS Neo that reduces browser context overhead, delegates navigation to Jev or other models, and reuses verified task experiences through OpenViking memory. Supports Cl…

    Python 1

  2. Agent-Watchtower Agent-Watchtower Public

    A passive, local observer of your Claude Code, Codex and Hermes usage: active hours, tokens, cost at API prices and response times, with OSRS-style levels, D&D attributes, quests and achievements. …

    TypeScript

  3. SocialCanvas SocialCanvas Public

    Replacement for Elfsight and other social widget services

    TypeScript

  4. Claude-Pocket-Checker Claude-Pocket-Checker Public

    A read-only macOS companion for regular Claude Desktop Chat: usage estimates and local draft model/effort advice.

    Swift

  5. Vibesomnia Vibesomnia Public

    Vibesomnia keeps your Mac awake and your AI workspace running—even with the lid closed. A macOS menu bar app with on-demand sessions, app-aware triggers, and controls built for remote AI developmen…

    Swift

  6. Claude-Browser-Accountant Claude-Browser-Accountant Public

    Local Claude.ai token estimates, API-equivalent cost and account-wide plan readings. Personal/work scopes stay separate.

    JavaScript