Platforms

The stack SPM runs on.

The platforms behind every engagement. Same stack we use ourselves — we don't sell what we don't operate.

Claude Code

Anthropic's official CLI for Claude. The dev environment SPM runs everything in.

· CLI · IDE · web · desktop
· Skills · MCP · subagents · hooks · plugins
· POM-based engagement model maps 1:1

Claude API

Direct API access to Opus, Sonnet, and Haiku — the engine inside every workflow.

· Tool use · prompt caching · batching
· Long context, vision, structured outputs
· SPM markup adds £1 / MTok wrapped in our SLA

Claude Managed Agents

Anthropic-hosted long-running agents — no infra to operate, billed by uptime.

· Per-agent identity & memory
· Runs your workflows on a schedule or trigger
· From £1 / agent-hour via SPM

AppAI cockpit

Mat's own multi-app control plane. PBI → Build → Preview → Push → Deploy for every app in the fleet.

· Registry-driven Next.js cockpit
· Live dev status per app
· Build · push · deploy receipts

ABC stack

The ~30-skill Mat Siems family that turns an idea into an intake, sector, proposal, app, diagram, and content pack.

· Skill-first composition
· Editorial-quality artifacts
· Tuned for SMB & enterprise pitches

Mat Siems

The operator. One person running SPM, AIBO, ABC, and the app fleet end-to-end.

· Engagement model: POMs + Managed Agents
· Stack hand-off ready for client teams
· Brand-aligned across every artifact

Shape of the stack

Four layers, stacked.

From the model up to ops. Anything in your stack should map to one of these — if it doesn't, it's probably a tool, not a platform.

Model layer

Claude Opus / Sonnet / Haiku via the Anthropic API. Long context, prompt caching, tool use, vision, structured outputs.

Claude APIClaude CodeManaged Agents

Skill layer

Invokable workflows that bundle prompt + tools + memory. The unit of reuse. ABC skills, codebase skills, persona skills.

ABC skills (~30)Codebase master skillsPersonAI personas

App layer

The 13 sites under ~/APPS/ plus the AppAI cockpit. Same Next.js 16 stack, same shadcn/tailwind v4 base, same deploy pipeline.

AppAI cockpit13 site reposAIBO / AIOS

Ops layer

What keeps it running: Supabase for auth/data, MongoDB for skill state, Vercel for hosting, GitHub for source of truth.

SupabaseMongoDB AIDBVercelGitHub

Live right now

What's running in production.

We don't sell what we don't run. Every service below is live under active production load.

Service	Stack	Status	Notes
AppAI cockpit	Next.js 15	Live	Orchestrates 13 site repos
Claude API	claude-sonnet-4-6	Live	Primary model for all agents
Supabase	Postgres + Auth	Live	Single-user gate + CMS store
Vercel	Edge + serverless	Live	Hosts all 13 V2 sites
MongoDB AIDB	Atlas M0	Live	Skill state + daily workflow log
GitHub	flexappdev org	Live	13 site repos + cockpit

Operating principles

Why we picked this stack.

We don't sell what we don't run

Every platform on this page powers production work on Mat's own 13 sites. If it broke for you, it would break for us first.

Local-first by default

Everything that can run on your laptop, does. Cloud is opt-in, not the default. Your IP doesn't have to live on someone else's server.

No lock-in we wouldn't accept

Your skills, agents, and PBIs live in plain files. If you sack us tomorrow, you walk away with everything that matters.

How we price it Agent surface

What runs on this stack

The fleet. Thirteen live sites.

Every site below ships from the same skill family — same auth gate, same admin shell, same backoffice. They're not bespoke; they're instances. If your project needs one of these patterns, we can fork it.

fad

Holding co — diagrams, agents, admin.

Personal portfolio — auth, legal, security headers.

yb100

Top 100 bunnies — Mongo + GA4 + GDPR consent.

Funky Success — 1,900+ non-fiction books, Stripe paywall.

Scroller / SP — Mongo-primary catalog + admin.

xmas

Top 100 Christmas — Runware images, cron pipeline.

wbp

World boroughs — Wikivoyage crawler, Leaflet map.

ybl