This project addresses the limitations of single-agent AI systems within the OpenClaw framework by introducing OpenCrew, a multi-agent operating system. It transforms a single OpenClaw instance into a collaborative AI team with specialized roles, enabling efficient task delegation, stable iteration, and automated knowledge accumulation. OpenCrew is designed for decision-makers and power users seeking to manage and scale AI team capabilities with minimal manual intervention.

How It Works

OpenCrew structures AI agents into three distinct layers: Intent Alignment (User + CoS), Execution (CTO, Builder, CIO, Research), and System Maintenance (KO, Ops). This multi-agent approach, where "channels/groups = positions" and "threads = tasks," prevents context inflation and isolates domain expertise. Core mechanisms include the Autonomy Ladder (L0-L3) for controlled decision-making, a task classification system (QAPS) for differentiated handling, and a two-step Agent-to-Agent (A2A) communication protocol leveraging Slack's Socket Mode for inter-agent coordination. Knowledge is systematically layered from raw dialogue to structured summaries (Closeouts) and abstracted reusable knowledge.

Quick Start & Requirements

To get started, users require a functional OpenClaw instance already integrated with a supported platform: Slack, Feishu, or Discord. Deployment is streamlined via a specific prompt sent to an existing OpenClaw agent, instructing it to clone the OpenCrew repository and merge necessary configurations. Alternatively, manual deployment steps are detailed in DEPLOY.md. Slack and Discord offer full support for thread-based task isolation, while Feishu's thread support is currently limited. The project claims a "10 minutes to get started" setup time.

Highlighted Details

• Multi-agent domain separation with dedicated channels/groups for each role.
• Robust Agent-to-Agent (A2A) collaboration protocol for seamless task delegation and execution.
• Mandatory structured outputs (Closeout/Checkpoint) ensure consistent deliverable quality.
• Autonomy Ladder (L0-L3) provides fine-grained control over agent decision-making authority.
• Layered knowledge accumulation mechanism transforms conversational data into organizational assets.
• Ops Review governance loop and KO knowledge extraction enhance system health and reusability.

Maintenance & Community

The project welcomes contributions via Issues and Pull Requests, particularly for architectural improvements, knowledge system enhancements, platform adaptations, and documentation. While specific community channels like Discord or Slack are not explicitly linked in the README, the project encourages collaborative development.

Licensing & Compatibility

OpenCrew is released under the MIT license, which permits broad usage, including commercial applications and integration into closed-source projects, with minimal restrictions.

Limitations & Caveats

Feishu platform users will experience limited thread support for task isolation due to underlying OpenClaw plugin constraints. The initial A2A setup process involves configuration changes that trigger OpenClaw gateway restarts, temporarily interrupting active agent sessions. An experimental "v2-lite" architecture is under exploration for a potentially lighter system.