ChatDev (DevAll)

Overview

ChatDev (branded as DevAll in v2.0) is an open-source platform for building and running multi-agent systems powered by large language models. Rather than requiring developers to write agent coordination code, DevAll provides a zero-code visual environment where users configure agents, nodes, workflows and tools; the platform orchestrates interactions and executes complex multi-step tasks.

Key features

• Zero-code workflow authoring: drag-and-drop visual canvas to design multi-agent topologies, define phases and role-based agents, and configure context flows.
• Multi-domain templates: out-of-the-box workflow templates for software development, data visualization, 3D generation (Blender), game dev, deep research and more.
• Web Console & real-time monitoring: start workflows from a browser, view logs, inspect intermediate artifacts, and provide human-in-the-loop feedback.
• Extensible runtime & SDKs: lightweight Python SDK for programmatic execution and automation; modular backend for adding new nodes, providers and tools.
• Scalable orchestration: supports different topologies (chain, MacNet DAG-style) and advanced collaboration paradigms developed in linked preprints.
• Safe execution options: Docker-based runtime modes and Git mode for version-controlled development.

Architecture & Components

• Backend: FastAPI-based server that manages runtime, agents, memory and tool execution.
• Orchestration layer: workflow engine that interprets YAML configurations to instantiate and sequence agent interactions.
• Frontend: Vite + Vue 3 Web Console for visual editing, launching and monitoring workflows.
• Extensibility: functions/ and runtime/ modules let developers add custom tools and providers.

Usage & Quick Start

1. Install prerequisites (Python 3.12+, Node 18+). 2. Sync backend dependencies and install frontend packages. 3. Run backend (e.g., uv run python server_main.py --port 6400 --reload) and frontend (VITE_API_BASE_URL=http://localhost:6400 npm run dev). 4. Use the Web Console to select templates, upload assets, enter prompts, and launch orchestrations. The Python SDK allows run_workflow(...) for automation.

Research & History

ChatDev grew from the ChatDev v1 virtual-software-company paradigm into DevAll (ChatDev 2.0), integrating several research ideas from related preprints (e.g., Experiential Co-Learning, MacNet, and multi-agent orchestration work). The repository links to associated arXiv preprints and provides branches for legacy and experimental features.

Notable dates

• Repository initial public availability: 2023-08-28.
• ChatDev preprint associated with the project was posted July 16, 2023 (arXiv:2307.07924).
• README announces the official release of ChatDev 2.0 (DevAll) on 2026-01-07.

Who should use it

• Product teams and researchers who want to prototype multi-agent workflows without heavy engineering.
• Developers seeking templates for automated software development pipelines, data analysis, or creative generation tasks.
• Researchers exploring LLM-based agent collaboration and orchestration strategies.