TrajOnco: a multi-agent framework for temporal reasoning over longitudinal EHR for multi-cancer early detection

TrajOnco is a research-oriented project (associated with a very recent paper) that applies agentic workflows to the high-stakes domain of oncology. While it demonstrates a sophisticated 'chain-of-agents' approach to solve the problem of long-context clinical history, its defensibility is low (3) because it relies on a 'training-free' methodology. This means its primary value lies in prompt engineering and architectural orchestration rather than a proprietary model or a massive, unique dataset. The quantitative signal (0 stars, 7 forks) indicates it is currently in the 'early peer review/researcher discovery' phase. Its biggest threat comes from platform giants like Google (Med-Gemini) and specialized medical AI firms like Tempus or Flatiron Health, who possess the direct EHR integrations and regulatory clearances required for deployment. As frontier labs improve the native long-context window and reasoning capabilities of models (e.g., Gemini 1.5 Pro's million-token window), the need for complex agentic memory management for longitudinal records may diminish, creating a high displacement risk within 1-2 years.