linke-ai/hermes-agent-team

Hermes Agent Team is a local web system that lets you run a small team of AI agents on your own computer, with each agent playing a specific role and working together to complete tasks. The project is built on top of the Hermes Agent framework and is a community experiment, not an official product from the framework's creators. All agent configurations and data stay on your local machine and are never sent to a remote server. The team is organized into two tiers. A Leader agent receives a task, breaks it into steps, assigns those steps to specialist agents, reviews the results, and produces a summary. Specialist agents can represent different functions: product management, software development, testing, design, or operations. Each agent has its own personality description, skills, memory, and tools. You write those descriptions yourself in plain text. A web interface lets you watch all of this in real time. You can see the conversation between agents, the tools each agent calls, the status of each subtask, and the final output. Agents communicate through a hub using two protocols (MCP and ACP), and task status is tracked on a Kanban board that the system manages automatically. The backend is Python-based, using Flask and an ASGI server. Storage is a local SQLite database. The frontend is plain HTML and JavaScript with no external framework. You start everything with a single command and open a browser to localhost. The system runs on Linux and macOS only. Windows users need to use WSL2. You must have the Hermes Agent CLI installed and configured before running this project. The README is written in Chinese. The project is released under the MIT license and maintained on a best-effort basis by a solo developer.