Orbit for Developers
One agent across your editor, browser, terminal, and docs — building its own context from every surface. The agent sees your running app.
Why this matters
Every AI coding tool connects the agent to one surface. Cursor connects it to a code editor. Claude Code connects it to a terminal. The agent writes code — but it can't see the result. You copy error logs, screenshot UI bugs, describe what went wrong.
Orbit gives the agent access to every surface you build with. The agent takes screenshots of your running app. It reads your terminal output. It reads your docs in the vault. It sees your code. You work in one window. The agent works across all of it.
What you get
Every surface is connected to the same agent.
Claude SDK with sub-agents, skills, plugins, and a built-in MCP creator. Configurable: plan-first or direct execution.
CodeMirror 6 with full LSP support. Syntax highlighting, code intelligence, file system access. Not VS Code — no extension ecosystem.
Embedded browser. The agent takes screenshots of your running app, navigates, clicks, scrolls, and fills forms autonomously. Click any element, describe the change.
Tabbed terminal with full agent access. Run commands, scripts, and manage processes.
Markdown editor for docs, PRDs, and notes. Write your requirements — the agent reads them as context while building.
Staging, commits, branches, merges, diffs, history, push and pull. All from within the environment.
What to know before you try it
Being upfront about tradeoffs.
The editor is CodeMirror 6, not VS Code. No VS Code extension ecosystem. Full LSP support, but a different tool.
No inline completions. No tab-autocomplete. All AI interaction is through the agent conversation. Different model for different workflows.
Claude-only for V1. Multi-model support (OpenAI, Google, local models) is on the roadmap.
macOS only. Apple Silicon. Windows and Linux coming.
Questions
How is this different from Cursor?
Cursor is a VS Code fork — you get the extension ecosystem and inline tab completions. Orbit is built from scratch (Tauri + React) with an embedded browser the agent can see and control. Different tradeoff: Cursor has more editor features, Orbit has more agent surfaces.
No inline completions?
No tab-autocomplete like Copilot or Cursor. All AI interaction is through the agent conversation. If inline completions are essential to your workflow, Cursor is the right tool. Orbit is agent-first, not autocomplete-first.
No VS Code extensions?
Correct. Orbit has its own skills and plugin system plus a built-in MCP creator. You won't get your favorite VS Code extensions, but you can build equivalent capabilities through skills.
Can I use my own API key?
Yes. Bring your own Anthropic API key — no token markup. Or sign in with any Claude account (even free tier). V1 is Claude-only. Multi-model support is on the roadmap.
Is my code private?
Your code stays on your machine. Orbit is a native desktop app — nothing is stored on our servers. AI requests go to Anthropic's API.
What platforms?
macOS (Apple Silicon) for V1. Windows and Linux are on the roadmap.
Free during early access
Sign in with your Claude account or bring your own API key. No setup required.