Project
48 specialist agents. 6 teams. One orchestrator that knows how to use all of them. Ember runs on Claude Code — you tell it what you want, it figures out who handles it and gets it done.
How it works
Ember isn't a single AI tool — it's a structured team of specialists. Six named teams cover every function: Build, Quality, Infrastructure, Data & AI, Research, and Content. Each team has a lead agent. The Orchestrator routes your request to the right lead, who sub-delegates to the right specialists. Work that can run in parallel does.
You don't prompt individual agents. You tell the Orchestrator what you want built, fixed, deployed, or written — and it handles the coordination. Skills compress multi-step procedures into a single command. MCP Integration Packs connect agents directly to live SaaS APIs — Stripe, Slack, Vercel, GitHub, and dozens more — so agents can take real action, not just generate code.
The result is something that feels less like a tool and more like having a competent team on call.
Build. Quality. Infrastructure. Data & AI. Research. Content. Each team has a lead agent and a roster of domain specialists. The Orchestrator routes to the lead — the lead runs the team. You deal with outcomes, not agent management.
Multiple agents run concurrently. Frontend and backend can build at the same time. Multiple projects can run in parallel sessions. The trade-off is token spend — the payoff is speed and throughput that changes what's possible in a day.
Skills are structured workflows invoked as slash commands — /deploy, /security-audit, /session-end. MCP Integration Packs connect agents to live SaaS APIs so they can take real action: create a Stripe subscription, post to Slack, deploy to Vercel. One install command, live in every agent.
Persistent memory across every session. Architecture decisions, context, and in-progress work carry forward. SESSION.md tracks what's active. HISTORY.md keeps the record. You never re-explain yourself.
How a request flows
You
Natural language instruction
Orchestrator
Breaks it down. Routes it.
6 Teams
Lead Agent (per team)
Sub-delegates to specialists
Specialists — run in parallel
frontend
backend
tester
devops
security
Real output. Real action.
In the wild
Ember isn't just running on my own infrastructure anymore. The first commercial integration is in progress — Ember being embedded inside a SaaS platform as the AI provider. Can't name it yet, but the build is in flight. That's the shift from framework to platform.
A SaaS platform (NDA)
In development — Ember embedded as AI provider for site generation, content editing, and SEO auditing
This website
Built and maintained by Ember agents
Sieno
AI-powered local SEO platform
bible-study-automation
Automated weekly Bible study content pipeline, 12-phase system
Snake Game →
Proof of concept, built in one session
Where it's heading
Ember started as a private framework I built for my own work. That phase is over. The first commercial integration is underway — Ember is being embedded inside a SaaS platform as the AI provider. The details aren't public yet, but the build is in flight.
The product roadmap is moving: CLI, agent teams, skills, MCP integration packs, and Ember Cloud are all in progress. Early access conversations are open for teams who want to build with agent-first architecture — properly — before it's widely available.
What's in the pipeline
Ember isn't waiting to be finished before it's useful. The first commercial integration is in progress, the CLI is being built, and the next capability layers are queued. Each one compounds the last.
Core framework
Live ✓CLI tooling
In progressSkills system
In progressMCP integration packs
In progressEmber Cloud
ComingCommercial platform
ComingCore framework
Live ✓CLI tooling
In progressSkills system
In progressMCP integration packs
In progressEmber Cloud
ComingCommercial platform
Coming"The best software isn't written by AI. It's written by humans and AI working in concert — with the right structure to make that actually work."
— Dave Bock
