Why we built Ember: from Cursor to Claude Code to our own editor
In shortWe built Ember because we wanted Claude Code as the agent, on the subscription we already paid for, inside an editor that kept the habits we had built up elsewhere: Tab, Cmd+K, Keep and Undo, and many agents in parallel. No existing product gave us both, so we built it for ourselves and then kept going.
Where we started
Our team was deeply invested in Cursor. Our rules, our shortcuts and our daily habits were all shaped by it: agent chat with Plan and Ask modes, several agents running at once, Tab to accept the next edit, Cmd+K to change a few lines, and Keep or Undo on every change.
Then it stopped being a viable answer for us, and we moved our work to Claude and Claude Code.
What the move cost us
Claude Code in a terminal is excellent. One agent in one terminal is easy to follow. The VS Code extension puts that agent beside your code, but it left a lot to be desired for the way we actually work. We missed next-edit predictions as we typed. We missed inline edit. We missed reviewing a change hunk by hunk before it landed.
Most of all, we missed being able to see everything at once. Five agents on five tickets, each with a pull request and CI, is a different job from one agent in one terminal. You need to know which branch is where, what each one is waiting for and what to do next, without reading five transcripts.
What we decided to build
We set two rules on day one that we have never relaxed. First, the agent is Claude Code as it is, on the subscription you already pay for. Ember never implements its own sign-in, never reads your credentials, and removes stray API keys from the agent's environment so nothing is billed per token by accident. Second, everything you configured for Claude Code keeps working: CLAUDE.md, skills, hooks, subagents, plugins, MCP servers and permissions are the same files the terminal reads.
Everything else is the editor we wanted around that agent: a board of every branch in flight, plan builds with several agents, a real review tab, Tab and Cmd+K, a codebase index, and a phone app for when you step away.
Why it is one app
Ember began as an extension. We changed course and made it one app, because some of what we wanted can't be done through an extension API: showing the lines Claude removed inline above a change, Keep and Undo buttons on each change in the editor, workspace tabs in the title bar with each branch's state, and asking before you quit while chats are still working.
If you are coming from Cursor or another VS Code-based editor, your habits carry over, and Ember imports your settings, keyboard shortcuts, snippets and extensions. What is different is underneath.