FleetQ: Open-Source AI Agent Orchestration Platform, Building a Self-Hosted Multi-Agent Task Control Center

FleetQ is a production-grade open-source AI agent orchestration platform designed to address key challenges in multi-agent collaborative management. Unlike traditional Python code frameworks, it offers visual DAG workflow building, integration with over 450 MCP tools, human-machine collaborative approval, and multi-agent team collaboration capabilities. It supports self-hosted deployment and uses PHP/Laravel as its tech stack, providing developers and enterprises with a complete agent task control center.

Against the backdrop of rapid development in AI agent technology, multi-agent collaborative management has become a core challenge. FleetQ was developed and open-sourced by escapeboy under the AGPLv3 license, supporting both self-hosted and managed cloud service modes. Its tech stack choice of PHP/Laravel stands out in the AI tool domain, leveraging Laravel's mature ecosystem and enterprise-level features, positioning itself as a production-grade platform rather than a mere code framework.

FleetQ's core concept system includes:

• Agent: The basic unit, with configurations such as role, goal, and skills, supporting LLM failover
• Skill: Reusable capability module to avoid duplicate configurations
• Experiment: Stateful execution process with a 20-stage pipeline
• Team: Multi-agent collaboration unit supporting 7 collaboration modes
• Workflow: Visual DAG template supporting cross-experiment reuse
• Project: Task container supporting Cron scheduling and budget control
• Signal: Event trigger source (Webhook/RSS, etc.)
• MCP Tool: Bridge for interaction with external AI systems These concepts form a complete agent orchestration language.

1. Visual DAG Workflow: Drag-and-drop to build DAGs with 8 types of nodes including agents, conditions, and manual tasks, enabling complex logic with zero code
2. Multi-Agent Team Collaboration: Supports 7 modes such as sequential, parallel, hierarchical, self-claiming, and adversarial debate, adapting to different scenarios
3. Human-Machine Collaboration Mechanism: Can insert manual task nodes to pause the process and wait for human approval/input, suitable for high-risk decision-making, quality control, and other scenarios Each mode supports weighted QA scoring and cross-validation to ensure controllable output quality.

• MCP Ecosystem Integration: Provides over 450 tools via the MCP protocol, supporting calls to external systems like Claude/Cursor
• Budget Control: Built-in credit ledger system based on actual API cost control; automatically pauses when exceeding budget
• Agent Evolution: Analyzes execution history and proactively proposes configuration optimization suggestions to achieve a self-improvement loop
• Production-Grade Features: Tenant isolation, encrypted credentials, distributed tracing, multi-tenant support, BYOK (Bring Your Own Key) compatibility, and local LLM compatibility
• Quick Start: 14 pre-built templates covering 5 major categories, with an AI assistant sidebar to reduce the learning curve.

With its unique positioning and rich features, FleetQ provides a new option in the field of AI agent orchestration. It demonstrates the vitality of PHP/Laravel in the AI era, lowering the application threshold through visual orchestration and human-machine collaboration. It is suitable for teams that value data sovereignty, seek production-grade solutions, and hope to reduce AI development costs. As the MCP ecosystem matures, FleetQ is expected to become an important hub connecting various AI capabilities.