Tycen McCann
Enterprise AI Solutions Leader & Indie App Developer
About Me
River guide. Seismic engineer. Disaster responder. M&A consultant. AI developer. I've never been great at picking one thing and sticking with it — but I've been pretty good at mastering the moment. Right now that's enterprise AI and unleashing decades of side quests, but ask me again in five years. I can't wait to see what comes next.
Professional Experience
- River Guide: 50+ trips through the Grand Canyon whitewater — paid fun
- Seismic Engineer: Designed highways, bridges, and hospitals. Moved to Haiti after the 2010 earthquake, assessed 420,000 structures, trained local engineers in seismic design, and helped rebuild 16,000 homes
- M&A Consultant: Full cycle — pre-deal diligence, integration, carve-out, post-deal optimization across four continents in fintech, manufacturing, and data platforms
- Cloud & AI: Got poached by AWS to help Fortune 500s adopt cloud. When GenAI hit, went all in — now I design AI adoption programs, run developer workshops, and architect production agentic deployments
- Builder: 6 apps live on the App Store, a cloud-hosted AI assistant, and a habit of shipping things on nights and weekends that won't quit
- Three languages: English, Spanish, Haitian Creole — four if you count prompting. YOU MUST...
For Fun
I build things on nights and weekends because I can't not build things - it's a problem. Six apps on the App Store, a personal AI assistant that runs my life (iykyk), and whatever I'm obsessing over this week. Scroll down to see what happens when your tools finally catch up with your ideas.
a bit cringe - but fun
Projects
Open Source
AgentCore Hub
Because a customer had a fleet of agents running in production and no real way to see what any of them were actually doing — and because when I get an excuse to burn through tokens, I let it cook.
An open-source control plane for Amazon Bedrock AgentCore — run all your agents and workflows from one place. It discovers every agent you've deployed automatically, with nothing to register or wire up by hand, then gives you a metrics dashboard, a chat playground, full OTEL execution traces, session history, an autonomous dev workflow, and remote Claude Code + Codex. Clone it, point it at your AWS credentials, and it just works — a small always-on core plus bolt-on modules you add only if you want them.
- Observe & operate: dashboard of real CloudWatch + OTEL metrics, an agent playground with live chat, step-by-step execution traces, and full session history
- Build & ship: chat-based agent builder, a 14-agent autonomous pipeline that turns a feature request into a PR, and remote/cloud Claude Code + Codex you drive from the browser
- Stack: Next.js/TypeScript, modular core + bolt-on modules; ships a Claude Code plugin —
/agentcore-hub:setupstands it up in one guided session
View on GitHub
Ember
Because I wanted to fire off a task, close the laptop, and go to my kid's game — without handing my code to someone else's cloud to do it.
Keep your session warm. An open-source coding agent — Claude Code and Codex — that runs inside your own AWS account on a warm micro-VM. Fire a task at your desk, close the laptop, reopen on the web or your phone, and the work is exactly where you left it. Your code, your keys, your bill, your model — nothing leaves your account.
- Runtime: AgentCore micro-VM with a persistent EFS workspace and warm/idle/cold session state
- Cost: bring your own Claude/ChatGPT plan for $0 marginal LLM cost — or Bedrock pay-per-token, fully in-VPC
- Deploy: open-core (MIT), one-command self-host with an account guard
codebuild-ios-mcp
Because a customer's AI agents could write iOS code but had no way to actually build and test it on a Mac. And because tokens ;)
An AWS CodeBuild macOS runner that builds and tests iOS apps, exposed to AI agents as MCP tools through a Bedrock AgentCore Gateway. An agent calls ios_test(branch, scheme) and gets back a structured pass/fail summary, per-test failures, presigned artifact URLs (xcresult, screenshots, video), and a logs link. Fully async — it returns a build id immediately, so calls never time out no matter how long the build runs.
- Compute: reserved MAC_ARM (Apple M2) CodeBuild fleets, with a size override for heavier builds
- Tools: seven MCP tools — test, status, logs, schemes, list builds, cancel — over an AgentCore Gateway
- IaC: CDK v2 (TypeScript) — one
cdk deployplus one gateway-wiring script
Live on the App Store
Easy Meal Planner
Because my wife and I would rather be mountain biking than figuring out what's for dinner & groceries.
Weekly meal planning with 200+ curated recipes, smart side pairings, and one-tap grocery ordering through all major grocers. Pick your meals, the app builds the shopping list, and you can order everything from the trails. Household sharing so we're both on the same page. Voice assistant if you need hands-free.
- iOS: Swift, SwiftUI, NovaSonic-Swift-SDK
- Backend: AWS Amplify, AppSync, DynamoDB, Lambda, S3
- GenAI: AgentCore, Strands, Claude Sonnet/Opus via Bedrock, NovaSonic2 speech-to-speech
- Integrations: Instacart, Chicory, and 30+ Grocery Stores
I also MCP'd it — any AI assistant can now search the recipes and order groceries. No auth, no signup, just vibes and Instacart links.
Good Stories
Because my kids love when I tell them stories — and I wanted that to keep going even when I'm not there.
Audio-first stories for kids built on a Waldorf education framework — no Brothers Grimm trauma. Each story teaches something real: how to ride a bike, how to manage big feelings, how to be a good friend. Full AI pipeline handles research, story arcs, character development, and series planning. ElevenLabs narrates, and I personally read, proof, and edit every story before it ships.
- iOS: SwiftUI, AVFoundation
- Backend: AWS Amplify, S3, Lambda
- GenAI: Claude for story generation, ElevenLabs narration
Talk to Me
Because my wife's energy is a yoga retreat and mine is a monster truck rally.
AI-mediated voice conversations that help couples communicate difficult feelings constructively. It sounds a little weird, but it works.
- iOS: SwiftUI, AVFoundation audio
- Backend: AWS Amplify, Cognito auth, DynamoDB
- GenAI: ElevenLabs conversational AI, Claude
Brush-Up
Because I volunteered to ump my son's Little League games and had no idea what the infield fly rule was.
Real-time baseball rules assistant with flashcards for learning and a speech-to-speech AI you can ask questions on the fly. Runner on second, pop-up to the shortstop — does the infield fly rule apply? Brush-Up knows.
- iOS: SwiftUI, Core Data, NovaSonic-Swift-SDK
- Backend: AWS Amplify, DynamoDB
- GenAI: Claude for flashcard generation
Scam Buster
Because my mom got scammed by a fake FBI agent and "signed" an NDA so she couldn't tell anyone — including me.
Screenshot anything sketchy — emails, texts, links — and get instant AI analysis with web verification. Built it so she'd never get caught alone with a scammer again.
- iOS: SwiftUI, Share Extension, App Groups, NovaSonic-Swift-SDK
- Backend: AWS Lambda, API Gateway, DynamoDB
- GenAI: Claude for content analysis, DuckDuckGo verification
TheBudget
Because every budgeting app I tried was either too simple to be useful or became a part-time job.
An AI-first personal finance app built around envelope budgeting — connect your bank, and an AI agent categorizes every transaction, learns your patterns, and manages your money like a financial advisor who actually knows you. Voice or text chat, receipt scanning from your camera roll, and a nightly analysis pipeline that catches what you'd miss. Built for families who want real visibility without spreadsheet energy.
- iOS: SwiftUI, SwiftData, NovaSonic-Swift-SDK
- Backend: AWS CDK, Lambda, API Gateway, DynamoDB, S3
- GenAI: Strands Agents on AgentCore Runtime (chat, classification, briefings, 25+ tools, specialist agents: investment, tax, debt, retirement)
- Integrations: Plaid (bank & investment accounts, transaction sync), Share Extension (snap receipts from anywhere)
Featured
Hosted OpenClaw
Because I have way too much going on.
A personal assistant on steroids — an autonomous AI agent that runs 24/7 managing email, calendar, groceries, stock trading, marketing content, and development tasks across every project on this page. It delegates to specialized sub-agents, remembers everything, and works while I sleep. I built a full platform so anyone can spin up their own assistant in minutes.
- Infra: AWS Lightsail, Terraform, Lambda, API Gateway, DynamoDB, AgentCore (MCP Gateway + Memory)
- GenAI: Amazon Bedrock multi-model routing, multi-agent orchestration
- Integrations: Google Workspace (Gmail, Calendar, Drive), Telegram, WhatsApp, GitHub, AWS, Twilio, Browser automation, Instacart, Alpaca, +45 and growing
Juno
Because I had six apps to market and the social-media energy of a tired dad with a day job.
A native macOS creative studio for social. Brand-aware agents brainstorm, plan, generate, and edit reels, carousels, posts, and clips — then hand you the controls at the one step that actually matters. Multi-tenant, so every brand gets its own assets, voice, and campaigns. A 12-step write→review→revise pipeline with a human gate keeps quality in and slop out.
- macOS: SwiftUI, SwiftData, AVFoundation editing (macOS 15+)
- Backend: AgentCore Runtime (Strands agents) + AgentCore Memory, Lambda generation, Media Lake + Bedrock Knowledge Base search
- GenAI: Claude Opus via Bedrock for planning/production/review; Imagen, Veo/Grok, and ElevenLabs for image, video, and voice