Show HN: Platos – like Claude Managed Agents but open-source and self-hosted (github.com)

Platos has emerged as a noteworthy open-source alternative to hosted AI agent services like Claude Managed Agents and OpenAI Assistants. Designed to be easily deployable on personal infrastructure, it simplifies the management and operation of AI agents through a comprehensive agent runtime packaged in a single Docker Compose setup. Key features include a streaming chat runtime, durable execution layer built on trigger.dev, and a universal MCP gateway that integrates multiple tool families behind one endpoint. This allows for a unified, secure architecture that is also capable of multi-tenancy, making it suitable for both SaaS applications and internal platforms. The significance of Platos lies in its empowerment of organizations to gain full ownership of their AI infrastructure, data, and model choices without vendor lock-in. Users can seamlessly integrate with leading LLM providers such as OpenAI and Anthropic, while maintaining control over API keys and sensitive data. The framework not only caters to developers by offering a robust SDK in multiple programming languages but also maintains crucial aspects of observability, data security, and scalability across deployments. The project emphasizes community involvement, inviting contributions and discussions, thus fostering an inclusive environment for ongoing improvements and shared knowledge in the AI/ML community.