proinsight-io/crewmeld

What it solves

CrewMeld is an enterprise AI digital employee platform designed to orchestrate, deploy, and manage AI agents. It solves the complexity of moving from simple chatbots to structured enterprise workflows by providing a unified system for managing digital employees, standard operating procedures (SOPs), and specialized tools, all while supporting on-premise deployment.

How it works

The platform uses a three-tier asset model:

• Digital Employees: Task execution entities with specific personas, LLM configurations, and bound tools/knowledge bases.
• SOPs: Visual workflows that orchestrate collaboration between digital and human employees, supporting conditional branching, scheduling, and human approval breakpoints.
• Tools: Atomic capabilities (JavaScript/Python) that run in isolated OpenSandbox containers to ensure platform stability.

It integrates a shared conversation runtime across multiple channels (WeCom, DingTalk, Feishu, Telegram, etc.) and uses RAGFlow for document parsing and hybrid retrieval to ground agent responses in company data.

Who it’s for

It is built for enterprises that need to automate complex, multi-role business processes using AI agents, as well as organizations requiring a secure, self-hosted AI workforce management system with human-in-the-loop oversight.

Highlights

• Visual SOP Orchestration: Drag-and-drop canvas for building multi-role collaboration flows with state machine validation.
• Isolated Tool Execution: Tools run in Docker/Kubernetes-based OpenSandbox containers to prevent system crashes.
• Multi-Channel Integration: Unified messaging interface supporting 8+ platforms including email and SMS.
• Enterprise-Grade RAG: Integrated RAGFlow for multi-format document parsing and hybrid search.
• Human-in-the-Loop: Built-in human approval nodes and a dedicated human-service workbench for handoffs.
• Broad LLM Support: Compatible with OpenAI, Anthropic, Google, DeepSeek, and local inference via Ollama or vLLM.