Melty — Chat-First Open Source Code Editor

Introduction

Melty is an open-source code editor built around the concept of chat-first development. Instead of writing code directly, you describe what you want in natural language, and the AI edits your files while you watch and approve changes. It aims to make AI pair programming feel native rather than bolted on.

What Melty Does

• Provides a conversational interface for describing code changes in plain English
• Streams AI-generated edits in real-time with diff visualization
• Maintains full context of your codebase for accurate modifications
• Integrates with git to track and revert AI-generated changes
• Supports multi-file edits from a single conversation turn

Architecture Overview

Melty is built on Electron with a TypeScript frontend. The editor component extends VS Code's Monaco editor for familiar editing ergonomics. AI interactions are handled through an LLM provider abstraction that supports multiple backends. The chat-to-code pipeline parses intent, gathers file context, generates a diff plan, and applies changes atomically.

Self-Hosting & Configuration

• Clone the repository and build with npm (Node.js 18+ required)
• Configure your LLM API key in the settings panel
• Set the workspace root to your project directory
• Adjust context window size and model selection in preferences
• Git integration works automatically when a .git directory is present

Key Features

• Real-time streaming of code changes as the AI generates them
• Diff-based approval workflow so you review before committing
• Codebase-aware context gathering for accurate edits
• Git integration for easy rollback of AI changes
• Open-source and self-hostable with no vendor lock-in

Comparison with Similar Tools

• Cursor — More mature and feature-rich, but proprietary and subscription-based
• Aider — Terminal-based; Melty provides a visual editor experience
• Continue — IDE extension approach; Melty is a standalone editor built around chat
• Cline — VS Code extension; Melty makes conversation the primary interface
• Windsurf — Similar AI-first concept but closed-source

FAQ

Q: Which LLM providers does Melty support? A: It supports OpenAI, Anthropic, and other providers via configurable API endpoints.

Q: Can I use Melty without an internet connection? A: You need access to an LLM API. Local models via Ollama or similar can be configured for offline use.

Q: Is Melty suitable for large codebases? A: It handles medium-sized projects well. Context window limitations of the underlying LLM may affect very large repositories.

Q: How does it differ from using ChatGPT for coding? A: Melty has direct access to your files, can edit them in-place, and maintains project context across conversations.

Sources

• https://github.com/meltylabs/melty
• https://melty.sh