Show HN: AutoICD API – AI clinical coding platform for ICD-10 and SNOMED
TL;DR
AutoICD is an AI platform that converts unstructured medical text into ICD-10 and SNOMED-CT codes, built for real clinical workflows.
Key Points
- Under the hood it runs a multi-layer ML architecture with custom-trained models and curated medical knowledge – not an LLM wrapper.
- SDKs exist for JavaScript and Python, plus an MCP server enabling integration with AI assistants.
- Use cases include automated coding from clinical notes, structured datasets for research, and scalable pipelines for healthcare data.
Nauti's Take
'Not an LLM wrapper' is exactly what you want to hear in a regulated context – but the proof is in the validation data, which is notably absent here. The combination of SDK and MCP server shows technical maturity: anyone building clinical AI pipelines today wants modular components, not monoliths.
The open question is whether AutoICD will tackle regulatory hurdles like FDA clearance or EU MDR, or stay comfortably in research and analytics territory – because that distinction determines whether tools like this ever touch real clinical decisions.
Context
Clinical coding is one of the biggest cost drivers in healthcare administration – errors cost billions and delay billing cycles. A reliable ML system that extracts ICD-10 and SNOMED codes directly from free text could fundamentally accelerate this process. The inclusion of an MCP server also positions AutoICD as a building block for AI agents in clinical environments, a space that is only just emerging.
Sources