Showcase
An AI brain you drop into any codebase. It crawls the repo, learns the architecture, and works like an employee — multi-agent, autonomous, mobile-controllable.
Generic AI assistants forget everything each session. myAI gives them persistent memory, cross-repo awareness, scheduled autonomous work, and human oversight from your phone.
62
Specialist agents
auto-discovered
136
Skill playbooks
trigger keywords
44
MCP gateway tools
shared fleet-wide
24
Managed repos
one hub
24/7
Off-hours runner
free-window work
One continuous cycle — from a sentence to a shipped, supported app.
Describe an app in plain English
Blueprint scaffolds a full prod-ready repo
Queue work for the off-hours runner
Agents build & ship autonomously
Drive it all from your phone
Capabilities shipping today — counts are live where the gateway and database are reachable.
Domain-owning agents auto-discovered by Claude Code — architecture, frontend, API, DB, security, ops, swarm coordination and more. Parallel lanes prevent collisions.
Repeatable, named procedures with trigger keywords — code review, OWASP audit, blueprint scaffold, productionise. Loaded on demand, zero idle token cost.
A local gateway exposing tools for memory, scheduling, repo orchestration, routing and recall — shared by every repo on the machine.
myAI is the first of three products. The others get their own showcase, done right.
myAI
The AI management framework — this page.
liveagentFlow
Idea→app 8-stage generation pipeline. See the full section below ↓
liveconnect
Bug & feature intake hub for shipped apps. See the full section below ↓
liveThe idea→app generation pipeline built on this framework. An 8-stage agent pipeline turns a plain-English idea into a working, tested application — real codegen, not scaffolding.
Describe the app in plain English
Solution-architect drafts the approach
Gap analysis against the stated goal
Business requirements captured
The bug & feature intake hub for shipped apps. A closed loop turns a user report into a deployed fix — AI-triaged, AI-drafted, human-reviewed.
A bug or feature request lands via the widget or portal
AI classifies severity, dedupes, and routes it
Fix-engine drafts a patch against the target repo
Change opens as a reviewable pull request
The master SHOWCASE.md, rendered live.
SHOWCASE.md. The dashboard reads it from the master repo mounted read-only at AI_ROOT — verify the volume mount in docker-compose.yml.Health alerts + bug/feature triage
This dashboard: what is running, what needs you, what is next — live across the whole fleet. Installable PWA, mobile-first.
A work queue an off-hours autonomous runner drains on the free model window — schedule a task, walk away, review the diff in the morning.
One keyword scaffolds a full Next.js + TypeScript + Tailwind + Mongo + Sentry repo wired to the framework — idea to production app.
Drive sessions from your phone — Telegram alerts and a CLI-mobile branch bridge keep CLI and mobile sessions perfectly in sync.
A single hub managing 24 repos — per-repo state, handoffs, health cards and READMEs, all self-contained and portable.
Tier-aware model routing maps each agent/task to the right model, prefers the free Fable window for batch work, and meters real token spend.
Technical requirements + contracts
UI/UX + architecture design pass
Real codegen — files land in the repo
Reviewed, tested, deployable
Run the pipeline fully autonomous with an AI project manager driving each stage, or take the wheel yourself and approve every stage transition.
Not scaffolding-only — the Build stage writes actual, working code into the generated project, not just planning documents.
Long-running stages execute as background jobs, so a generation run survives a closed tab and picks up where it left off.
A browsable, tree-view artifact viewer across every stage — inspect and download generated files without leaving the workspace.
Merged fix ships to the reporting app
Incoming reports are classified, deduplicated, and prioritized by AI before a human ever looks at the queue.
Drafts a real patch against the target repo for qualifying bugs — not just a ticket, a candidate fix.
A single dashboard for operators to review reports, approve fixes, and track the loop from intake to deploy.
180+ tests back the support→triage→fix→PR→deploy loop, so the automation is verified, not vibes-based.