Elsechord

Open infrastructure for reliable autonomous agents.
逆律成调 · 为可靠的自主智能构建开放基础设施

We are Elsechord, an independent team exploring how autonomous agents can investigate, collaborate, and act with clear boundaries. Our work begins in security operations, with a broader ambition: make consequential agent actions inspectable, controllable, and verifiable across domains.

Our name brings an alternative possibility into a shared chord. We value original ideas, precise engineering, and the care it takes to make complex systems understandable.

• Evidence before action. Decisions should connect back to inspectable observations.
• Explicit authority. Agents need defined permissions and meaningful human intervention.
• Verifiable outcomes. An action is only part of the story; its effects need independent checks.
• Reproducible work. Open code should come with runnable examples and evidence others can examine.

CyberGuard is our first project: an evidence-driven autonomous security operations team built on AgentTeams.

It connects specialist agents across investigation, response planning, human approval, controlled execution, and recovery verification. CyberGuard contributes the security tools, evidence model, response policies, reproducible scenarios, and audit console; AgentTeams provides the underlying orchestration and collaboration infrastructure.

The project is under active development. The published response executor currently runs in simulation mode; real mutating actions require additional integrations. Deployment requirements and current boundaries are documented in the project repository.

We welcome reproducibility reports, technical feedback, and contributions from people working on agent systems and security operations. Start with CyberGuard's issues to discuss a concrete problem or proposed improvement.

Built by Shuuyou and Armaygooser in Hangzhou, China.