Running team AI agents shouldn’t mean handing your data to another SaaS. Platypus is an open-source, self-hosted full-stack application for teams building AI agents on their own infrastructure and models. It helps you assemble agents that can reason, call tools, and run on schedules by bundling agent configuration, reusable skills, sandboxed execution, and workspace controls in one application. Key features: • Agents, skills & sub-agents – define an agent’s model, instructions, and tools, then add reusable skills and delegated sub-agents • MCP support – connect agents to local and remote data sources through Model Context Protocol • Isolated sandbox – provide shell and filesystem access in a per-workspace environment, with Docker and SSH reference backends • Schedules & webhooks – run agents on cron or one-off schedules and receive HMAC-signed callbacks with filtering and retries • Provider choice – use hosted, proxied, or local models through providers including OpenAI-compatible endpoints It’s open-source (MIT license). Link in the reply 👇