How to make multiple agents work together like an elite team — autonomously dividing tasks, communicating efficiently, and collaborating seamlessly?
The openJiuwen community released the latest version of JiuwenClaw, which adds support for AgentTeam — a multi-agent collaborative capability. It suggests that the next big leap after Harness Engineering will be Coordination Engineering.
In in-depth tests, this team collaboration mechanism has demonstrated remarkable stability —team members have clear roles, collaborate autonomously with seamless coordination, and the entire workflow requires no human intervention.
What is the real hardcore?
This device can automatically assemble itself “well-trained” team of agents — and with that team, it can produce a solid, logically rigorous 200‑page technical PPT in under 20 minutes.
Project links: https://github.com/openJiuwen-ai/jiuwenclaw
Testing JiuwenClaw “Coordination Engineering” In Action
You want to get deep insights, but you don’t have the time or energy? A 200‑page, content‑rich PPT in under 20 minutes.
In our trial, we asked it to conduct an in‑depth investigation of OpenClaw technology and break it down across 10 core aspects. It assigned an agent for each of the 10 core aspects. Each agent had to create 20 PPTs slides under one theme. Finally, the 10 sets of slides were merged into a complete, 200‑page technical presentation.
It took under 20 minutes to complete the entire process. The PPT that resulted was detailed, logically organized, and impressively effective.
JiuwenClaw AgentTeam: Technical Overview of Three Core Capabilities
AgentTeam’s core philosophy is simple: it simulates how teams work in real life.
- A Leader Agent is responsible for requirement analysis, team building, and task planning.
- Multiple Teammate Agents claim tasks, execute independently, report results, and collaborate through a shared workspace.
- Key milestones during execution require approval by the Leader, while fault recovery occurs automatically.
1. Hierarchical autonomous collaboration: The leader orchestrates intelligently while the team executes independently
JiuwenClaw agentTeam delegates this responsibility to Leader Agent.
What a leader does
- Dynamically builds the team: Assigns roles and members dynamically based on the goal. It can be used to add members or take them away on demand if more help is needed during the execution.
- Planning tasks: Decompose the objective into specific tasks and establish dependencies. “analysis can only start after data collection is complete”).
- Assigns and monitors: After creating tasks, it tracks progress in real time—who claimed what, who completed what, who ran into issues—and adjusts accordingly.
What teammates do
- Claim tasks proactively: Browse the task board and claim tasks that match their capabilities.
- Execute independently: Complete their work within their own workspace.
- Report Results: Inform the Leader of the result and the other dependents.
Team members drive the core workflow through task collaboration—claiming, executing, completing, unblocking downstream tasks—discussing plans, negotiating priorities, flagging issues, requesting support.
Both channels run in parallel, with task dependencies managed automatically—not simply mechanical distribution and aggregation.
2. Team Files: Shared team file space
JiuwenClaw AgentTeam solves this with Team Workspace—a team‑level shared file space that all members can transparently access. Every Teammate’s directory mounts automatically a shared path to the same workspace.
3. The Full Lifecycle of Management: Plan approval to automatic fault recovery
3.1 Leader Approval
AgentTeam provides a two‑layer approval mechanism:
- Plan mode: For important tasks, a Teammate first submits an execution plan for Leader approval.
- Approval of Tool The Leader must approve sensitive operations (e.g. deletion of files, call to external APIs or modification of shared configurations) by a Teammate.
3. Event‑Driven Mechanism
AgentTeam mitigates this with an event‑driven mechanism, using both external and internal events:
- External events: Task state changes, member lifecycle changes, inter‑member messages—any meaningful change triggers an event.
- Internal events: Framework‑generated self‑check events (mailbox polling, task board polling) act as a safety net.
Agents are awakened automatically after an event has been triggered (e.g. Teammates who have not claimed tasks or tasks that were timed out by the Leader).
There are three types of persistent teams:
With Persistent mode enabled, teams can be preserved across sessions: The next time you need the team, you can restore it with one click—create a new session space, restart the team members, and you’re ready to go, without rebuilding the team from scratch.
TeamMonitor Version 3.4
TeamMonitor providing observability in two dimensions:
- The Query API is a query-based interface. You can check the team’s information at any time, including member status, progress on tasks, etc.
- Event stream Subscribing to events of the team in real-time. Task completions, member state changes, messages sent/received… all events can be consumed one by one via an asynchronous iterator. From these events, you can create dashboards, log systems or external workflows. The team can audit and trace every step.
OpenJiuwen agentTeam architecture: the core underpinning
AgentTeam’s core principles can be summarized into three main points.
- A shared task list allows for consistent collaboration: all members have access to the same list of tasks. Each agent independently claims and executes tasks based on the team goal, task definitions, and its own capabilities—ensuring natural information consistency.
- Dual‑drive model of messages and tasks: Members drive the core workflow through task transitions, while also continuously discussing and negotiating via a message channel outside the task system—covering everything from structured execution to unstructured communication.
- Role and tool engineering: RolePolicy defines the behavioral norms and decision boundaries of the Leader and Teammates within the team. TeamTools provides team members with coordination abilities. It is the role that determines “what should be done,” The tools that determine “what can be done.”
JiuwenClaw
JiuwenClaw, a Chinese brand of sonic claws is available. “Claw” Agent developed on top of the openJiuwen open‑source community. It natively supports multi‑agent collaboration and agent self‑evolution. The core design philosophy is simple: Understand what you want, and evolve autonomously.
Beyond AgentTeam, JiuwenClaw is also very easy to install and deploy – a single command gets you up and running. For a quick start, refer to: https://github.com/openJiuwen-ai/jiuwenclaw/blob/develop/docs/en/Quickstart.md
In addition, JiuwenClaw offers several advantages in autonomous task planning, self‑evolution, context compression and offloading, browser manipulation, and overall “lobster‑like” handling:
- Autonomous task management: always ready when you are : JiuwenClaw features a task planning mode, which is essentially a to‑do list for the AI. The user can interrupt, add or modify tasks dynamically at any point.
- Self-Defense‑Learn to Develop Skills This proactive feature records execution errors, feedback and analyzes root causes, and then generates specific improvement suggestions. An evolution approval window then pops up for the user – every update is your call.
- Context compression & offloading : Reduce costs effectively by managing the context length.
- Layered memory: Retrieval of operation and scenario traces can be intelligently retrieved and stored for long periods.
- Browser manipulation: Takeover the entire browser environment by automatically accessing profile data such as cookies and cache.
OfficeClaw
OfficeClaw is the enterprise version of Huawei Cloud AgentArts. It integrates multi-agent collaborative task planning and tool invocation on Huawei Cloud AgentArts.
Join the Community & Explore openJiuwen
openJiuwen Download Links
JiuwenClaw Download Links
- JiuwenClaw on GitHub: https://github.com/openJiuwen-ai/jiuwenclaw
- JiuwenClaw on AtomGit: https://gitcode.com/openJiuwen/jiuwenclaw
- AgentArts on Huawei Cloud：https://www.huaweicloud.com/product/agentarts
- OfficeClaw on Huawei Cloud：https://www.huaweicloud.com/product/agentarts/officeclaw.html
Note: OpenJiuwen is to be commended for its resources, including images, videos and other details.
Michal Sutter, a data scientist with a master’s degree in data science from the University of Padova is an expert. Michal Sutter excels in transforming large datasets to actionable insight. He has a strong foundation in statistics, machine learning and data engineering.
