Report Abuse

Basic Information

ChatDev IDE is a toolkit and browser extension for designing, building and testing AI agents and interactive NPCs. It provides a visual environment that combines a game-like simulation (AI Town) for interacting with customizable roles, a conventional chat window for LLM conversations, and a Prompt IDE for creating prompt flows. The project emphasizes prompt engineering via editable Prompt Flow YAML files, JavaScript scripting inside nodes, and visualization of execution as a directed acyclic graph. It supports importing and exporting agent definitions and works with many LLMs and models. The repository includes build and installation instructions for using the extension from web stores, manual loading, or building from source with yarn. The main purpose is to let users assemble LLM calls, prompts, scripts and tools into reusable agent workflows and to iterate and share agent designs.

Links

Categorization

App Details

Features
Key features include a Game Window that simulates an AI Town where customizable NPCs can be controlled and interacted with, and a Chat Window that provides a standard LLM chat UI. The Prompt IDE offers a Prompt Flow Editor with YAML syntax highlighting, automatic completion for NPC names and prompts, dual-screen editing, and real-time visual updates. Prompt Flow visualization shows execution progress of nodes as a DAG. JavaScript support allows custom node logic and cooperation between NPCs. The project supports many LLMs and models including mainstream and open models, agent import/export for sharing, and multiple installation paths including web store, manual extension loading, and building from source. The repo also highlights TypeScript-based tooling and integration with a Promptflowx development suite.
Use Cases
This project helps developers, designers and researchers rapidly prototype and iterate on LLM-driven applications and agent behaviors. It accelerates prompt engineering by combining an editor for prompt flows, live visualization of execution, and JavaScript hooks for custom processing so complex interactions can be composed and debugged. The Game Window enables experiential testing of NPC roles and social simulations, while the Chat Window offers a familiar interface for single-session testing. Export and import make it easier to share agent designs or reuse flows across devices. Multi-model support allows testing on different LLM backends. The YAML-driven flow and DAG visualization improve observability of how prompts, tools and scripts connect, aiding debugging and reproducible development.

Please fill the required fields*