Hey Folks,
This project started because I simply wanted to build a new IDE to fix my own workflow bottlenecks. But I immediately hit a brick wall trying to build the tooling for it. I spent months, days, and nights trying to hack and fix VS Code extension webviews just to get a decent UI to render.
I realized I was fighting a 33 year old architectural problem: extension vendor lock in. If you want your dev tool to reach people today, you have to write Electron/TypeScript for VS Code/Cursor, and Kotlin/JVM for JetBrains.
So I stopped building the IDE, and I built the fix instead.
Meet OXP (Open eXtensions Protocol).
OXP is an open source universal standard that lets you write your extension once in React/WASM and run it natively across every major editor.
How I fixed the Webview problem:
This isn't a slow iframe hack. OXP uses a secure WebAssembly sandbox and a zero-latency IPC bridge. Your React code triggers an action, and OXP translates it to native IDE commands.
In VS Code, it binds directly to the native extension API.
In JetBrains, it uses JCEF to render as a native floating OS window.
You get blazing fast native speed from a single codebase.
**The Accidental MCP Fix:
While building this universal host layer, I realized OXP perfectly solves the current Model Context Protocol (MCP) configuration hell.
Instead of manually editing configurations for Cursor, Copilot, and JetBrains individually, OXP acts as a system level MCP router. If you run oxp install @modelcontextprotocol/server-postgres, the OXP daemon instantly wires that database context into the AI configurations of every detected IDE on your machine.
I'm opening up the infrastructure today. The CLI is live, and you can test it on your machine right now.
Initialize your environment: npm i -g @oxprotocol/cli
Architecture and Docs: https://oxp.sh
I'll be in the comments all day to talk about the WASM bridge, IPC latency, fighting with JCEF, and why extension silos need to die.
Top comments (0)