A native, AI-powered code editor for every platform. Written in TypeScript, compiled to native via Perry. No Electron. No compromise.
Built from scratch in TypeScript, compiled to native binaries via Perry. Every component is designed for performance, modularity, and developer joy.
No Electron, no Chromium, no V8 runtime. Perry compiles TypeScript directly to native binaries. Under 50MB, under 100MB RAM, sub-second startup.
AI isn't bolted on — it's woven into the editor's core. The AI sees your syntax tree, your errors, your git diff, your terminal. Deep context, not just text.
Claude, GPT, Gemini, Ollama, or your company's private model. Your API key, your choice. Mix different providers for different features. No vendor lock-in.
Describe what you want. The agent reads your code, plans changes, edits files, runs tests, and iterates — with per-hunk approval before anything is committed.
Review pull requests in-editor with full syntax highlighting, LSP intelligence in diffs, and AI-powered annotations that catch bugs before humans do.
hone-editor, hone-terminal, and hone-core are independent packages. Use the editor component in your own app. Build your own IDE on Hone's foundation.
When you eliminate the browser, everything gets faster. These are our targets — and we intend to hit every one.
Hone never resells API access. You bring your own keys and route different providers to different features. When the next hot model drops, it works immediately.
Context-aware suggestions powered by any provider. Sees your syntax tree, not just raw text. Renders at native speed with zero Electron overhead.
The AI automatically sees your open files, errors, terminal output, and git state. No manual @file tagging. Switch models mid-conversation.
Multi-file edits, terminal commands, test execution, error recovery. Full transparency with a live activity log and diff approval before any change lands.
AI pre-analyzes every diff for bugs, security issues, and style problems. Review annotations appear inline. One-click fixes. Submit reviews back to GitHub/GitLab.
Route each AI feature to a different provider. Use a fast local model for autocomplete, a powerful cloud model for agent tasks, and something in between for chat. Your infrastructure, your budget, your rules.
Written once in TypeScript. Compiled to native on each platform by Perry. Native look, native speed, everywhere.
Hone is in early development. Follow along, contribute, or just watch.