OpenMed is a local-first healthcare AI platform that performs clinical named entity recognition and HIPAA-compliant PII de-identification entirely on-device.
The tool addresses the need to extract structured clinical insights from medical text while ensuring patient data never leaves the user's network. It works by running inference locally on the user's hardware after model artifacts are downloaded, keeping all processing on-premise. The platform supports multiple deployment targets including Apple Silicon via MLX, Android via ONNX Runtime Mobile, and browsers via Transformers.js, with a Python SDK as its core interface.
Organizations handling sensitive patient information should consider this tool if they operate under strict data residency requirements or cannot route clinical data through cloud services. It suits healthcare providers, research institutions, and compliance-sensitive deployments that need to maintain full control over where computation occurs. The project provides a large catalog of pre-built models covering clinical NER and de-identification across multiple languages, reducing the need to train custom models from scratch. The README does not make direct comparisons to alternative solutions.
The project maintains active development across multiple platforms and languages, with dedicated implementations for iOS via OpenMedKit and Android export pathways. Documentation covers both consumer and agent-based usage patterns, including MCP server integration and a command-line interface. The tool registry and skills catalog indicate ongoing expansion of callable interfaces and pre-built workflows for common healthcare NLP tasks.
