Welcome to the official DevEx roadmap! This document outlines the upcoming features and improvements weβre working on or planning to build. Contributions and suggestions are highly welcome β check out our GitHub repo to get involved.
π‘ DevEx is an open-source, cloud-based IDE like Replit β but built for hackers, contributors, and learners.
β Current Focus
These are the features that are either being built or planned for the near future:
π§ 1. MCP Server β AI Assistant in the REPL
Level: Medium
Goal: Enable users to interact with their REPL session using an AI assistant (e.g., Gemini, Claude).
Overview:
β’We'll introduce a Model Context Protocol (MCP) server that runs alongside each REPL session in the Kubernetes pod.
β’It will communicate with the REPL through internal APIs, allowing the assistant to access files, terminals, and logs.
β’Shutdown Manager will also account for AI-activity (preventing idle shutdowns if AI is in use).
β’Web UI and API documentation will be provided for full integration.
π» 2. DevEx CLI + SSH Access
Level: Medium
Goal: Give users the power to interact with their REPL directly from their terminal.
Overview:
β’A lightweight DevEx CLI will allow users to create, manage, and connect to REPL sessions.
β’SSH access will enable full terminal control β great for advanced workflows.
β’The shutdown manager will be updated to handle active SSH sessions.
β’Will include secure key management, SSH hardening, and simple installation docs.
π 3. GitHub Integration for Project Bootstrapping
Level: Easy
Goal: Let users start REPL sessions from their own GitHub repos instead of templates.
Overview:
β’Migrate from OAuth to GitHub Apps for authentication.
β’Add UI and API flow to import a GitHub repo directly into the REPL session.
β’Use GitHub API to clone and sync projects.
π§ͺ 4. AI Code Sandboxing (Experimental)
Level: Medium
Goal: Safely test AI-generated code inside REPLs before integrating into real apps.
Overview:
β’This feature is distinct from the MCP server. While MCP provides assistant interaction, sandboxing focuses on previewing AI-made changes.
β’AI will edit a session sandbox; users can then test the app live via a shareable preview URL.
β’If happy with the changes, users can manually merge or discard them.
β’This ensures experimentation without compromising your main app logic.
π Future Ideas (Not Yet Prioritized)
These are potential features weβd love to explore in the future:
β’Multi-user real-time collaboration
β’Persistent filesystem (bind storage volumes)
β’Workspace Templates from community
β’Plugin system for language/runtimes
β’Built-in secrets manager for REPLs
π Want to Contribute?
Weβd love your help β from fixing bugs to building features or designing templates!
Check out: