Large language model (LLM)‑based multi‑agent systems (MAS) have demonstrated strong potential for tackling complex tasks, yet their development faces a trade‑off: code‑centric frameworks are expressive but engineering‑heavy, while no‑code builders are user‑friendly but restrict agent interactions to author‑defined workflows.
ChatDev 2.0 (referred to as DevAll) is a no‑code platform that aims to deliver both high expressiveness and ease of use. For expressiveness, DevAll pairs a declarative executable‑graph abstraction with a cycle‑aware execution engine, allowing heterogeneous agents and dynamic, cyclic interactions to be represented and executed within a single framework. For usability, an integrated visual interface lets users author, run, monitor, and inspect MAS—including human‑in‑the‑loop steps—without writing any code.
Experiments show that DevAll reproduces state‑of‑the‑art MAS on three representative tasks with competitive performance, all without task‑specific orchestration code, highlighting its effectiveness as a general‑purpose LLM‑based MAS platform. The project is open‑source at https://github.com/OpenBMB/ChatDev.
Review: By marrying visual tooling with a graph‑based execution model, DevAll lowers the barrier to MAS development while preserving the ability to model complex interactions, offering a reusable benchmark for future LLM multi‑agent research.