Summary

CrewMeld is an Enterprise AI Digital Workforce Platform designed for managing AI agents. It provides visual orchestration of complex workflows (SOPs), supports a wide array of LLM providers and messaging channels, integrates knowledge bases via RAG, and offers full private deployment capabilities, enabling businesses to deploy and manage AI employees akin to human team members.

How It Works

The platform centers on three core asset types: Digital Employees (identity, tools, LLM/KB bindings), SOPs (visual, multi-role collaboration flows with step-level orchestration, human approval, and breakpoint resume), and Tools (atomic capabilities run in isolated OpenSandbox containers). Its five-layer architecture spans Trigger, SOP Orchestration, Digital Employee, Tool execution, and External Systems integration, built on a robust tech stack including Next.js, Bun, TypeScript, PostgreSQL, and Redis.

Quick Start & Requirements

Installation is streamlined via Docker Compose using ./start.sh (or start.bat/.\start.ps1 on Windows), which handles environment setup and secret generation. Local development requires Bun (1.3.9+) and Docker. Key dependencies include PostgreSQL 17, Redis 7, MinIO, RAGFlow (v0.23.1), and OpenSandbox. Minimum idle memory is ~500MB, scaling to ~7GB with RAGFlow and OpenSandbox. Official resources include the website (https://crewmeld.ai/) and user manual (https://proinsight.gitbook.io/crewmeld).

Highlighted Details

• LLM Agnosticism: Supports 13+ LLM providers (OpenAI, Anthropic, Google, Tongyi Qianwen, DeepSeek, ERNIE, Hunyuan, Moonshot, Zhipu, Doubao, MiniMax, Ollama, vLLM), including specialized coding models.
• Omnichannel Integration: Connects via 8+ messaging platforms (WeCom, DingTalk, Feishu, WeChat OA, Email, SMS, Telegram, Discord) using a unified plugin system.
• RAG Knowledge Base: Integrates RAGFlow (v0.23.1) for multi-format document ingestion, OCR, and hybrid retrieval (vector + BM25).
• Isolated Tool Execution: Utilizes OpenSandbox (Docker/Kubernetes) for secure, isolated execution of JavaScript and Python tools, ensuring platform stability.
• Flexible Deployment: Offers Docker Compose for simpler setups and Helm charts for production Kubernetes deployments, supporting on-premise installations.

Maintenance & Community

The provided README does not detail specific community channels (e.g., Discord, Slack), active maintainers, or a public roadmap.

Licensing & Compatibility

Released under a modified Apache License 2.0. Commercial use is permitted for backend services and enterprise development platforms, but requires a separate license for multi-tenant services (defined as one tenant per workspace) and prohibits altering frontend branding. Contributor code is explicitly permitted for commercial use.

Limitations & Caveats

The modified license imposes specific commercial restrictions, particularly around multi-tenancy and frontend branding, necessitating careful review for enterprise adoption. The README does not detail known bugs, alpha/beta status, or platform limitations beyond these licensing terms.