HiClaw - Alibaba Cloud's open-source multi-agent team collaboration system

HiClaw is Alibaba’s open source Agent team collaboration system, positioned as “Team version” OpenClaw ”. The system introduces Manager Agent as an AI steward to automatically coordinate multiple Worker Agents to complete complex tasks. The core highlights of HiClaw include that Workers do not hold real credentials (only Consumer Token), and are securely isolated and leak-proof; the built-in Matrix server allows you to check and intervene at any time on your mobile phone; you can create Workers conversationally, automatically dispatch orders, and monitor progress. The system is suitable for independent developers and one-person companies to quickly form an AI employee team to achieve “one person commanding multiple Agent collaboration”.

HiClaw’s main features

• Manager-Agent coordination : Manager Agent, as an AI steward, automatically creates Workers, assigns tasks and monitors execution progress through natural language dialogue, eliminating the need to manually configure each Agent.
• Multi-Agent collaboration : Supports the creation of multiple Worker Agents (such as front-end, back-end, product) by role. Each Worker has an independent skill set and memory space to avoid task confusion and context pollution.
• Security Credential Management : The Worker only holds the Consumer Token issued by Higress, and sensitive credentials such as the real API Key and GitHub PAT are stored uniformly in the AI gateway, so that even if the Worker is compromised, the core credentials cannot be leaked.
• Built-in IM system : Integrates the Tuwunel Matrix server and Element Web client, eliminating the need to connect to Feishu or DingTalk. It supports real-time communication between browsers and mobile devices out of the box.
• manual supervision intervention : All Agent communication occurs in the Matrix group chat room. Administrators can observe the conversation process at any time and directly @ any Agent to intervene or correct.
• Dynamic skill loading : Worker has built-in find-skills capability, which can actively retrieve and install skills.sh community skills, and supports switching to a private skills library to meet enterprise customization needs.
• file sharing storage : Using MinIO as a centralized file system, intermediate products such as codes and documents are transmitted through file sharing instead of group chat, keeping the conversation context streamlined.
• One-click deployment and operation : A single command completes the complete installation of Higress gateway, Matrix server, MinIO storage, Manager and Web client, supporting in-place upgrade and data persistence.

HiClaw’s technical principles

• Layered architecture design : HiClaw adopts a four-tier architecture to achieve separation of responsibilities. The user layer interacts with the system through the Matrix protocol; the coordination layer has the Manager Agent responsible for task disassembly and Worker life cycle management; the execution layer deploys stateless Worker Agent containers; the gateway layer uses the Higress AI Gateway to uniformly proxy all external service access and host real credentials. Layering ensures that workers cannot access sensitive information even if they are attacked, and it also facilitates independent expansion of components in each layer.
• Security certificate transfer mechanism : The Worker obtains the Consumer Token from the Manager when it starts. The token only identifies the identity but has no actual permissions. When a Worker calls LLM or GitHub API, the request first passes through the Higress Gateway. The gateway injects the corresponding real API Key based on the Token identity and then forwards it to the upstream service. The design achieves complete isolation of credentials. The Manager knows the task content but cannot access the key. The Worker executes the task but cannot export the credentials, thus eliminating the risk of credential leakage from the architecture.
• Anti-panic group and message routing : When realizing multi-Agent collaboration based on Matrix group chat, HiClaw designs a precise wake-up mechanism. Group chat messages do not trigger any Agent by default. Only when the message content @ a specific Agent or contains specific keywords, the corresponding Agent will call LLM to respond. This avoids the cost and waste of all Agents reasoning about each message, while maintaining the coherence of the group chat context, making the communication overhead controllable when multiple people and multiple Agents collaborate.
• Hierarchical storage and state management : The system uses a dual storage strategy to separate different types of data. The Matrix server only retains conversation messages and decision records as the Agent’s “memory” context; MinIO object storage is responsible for saving large-volume intermediate products such as code files, documents, and temporary data. The Worker itself is designed to be stateless. Files are accessed by mounting shared volumes and can be destroyed after the task is completed. This not only controls the length of the LLM context to reduce costs, but also enables quick start, stop, and elastic scaling of the Worker.

HiClaw project address

• Project official website ：https://higress.ai/hiclaw
• GitHub repository ：https://github.com/higress-group/hiclaw

HiClaw application scenarios

• Full stack development for independent developers : A single developer can quickly form a virtual team and create multiple workers for front-end, back-end, testing, etc. to develop in parallel. The Manager automatically coordinates dependencies to achieve the efficiency of “one person supporting a team”, which is suitable for independent package acceptance or side projects.
• One-person company MVP quick verification : Entrepreneurs describe product requirements through natural language, and the Manager automatically breaks down into subtasks such as PRD writing, technology selection, core development, and release material preparation, etc., and assigns them to different role workers for execution, completing the complete process from idea to launch within a few days.
• Enterprise Digital Workforce Pilot : Enterprises can deploy HiClaw locally and privately, and hand over repetitive tasks (such as data report generation, code review, and document organization) to the Agent team. The human employees focus on decision-making and innovation, and gradually explore the human-machine collaboration model.
• Open source project maintenance : Open source maintainers create Workers to handle daily tasks, such as automatically responding to issues, generating changelogs, reviewing PRs, and synchronizing document translations. Managers monitor task status and remind maintainers to intervene when exceptions occur. ©