FreshContext's current home is https://freshcontext.dev/ . This is a progress update to the existing product, not a new launch.
A fixed-time local check of Core 0.5.2 evaluated five synthetic inputs: a recent dated record, a historical record, an undated record, a failed retrieval, and a recent-looking record with an unknown source.
What should happen to retrieved context before it reaches the model?
Most RAG and agent systems focus on retrieving more relevant material. That matters, but relevance alone does not tell an AI system whether a source should be trusted, cited, refreshed, verified, used as background, watched, or excluded.
FreshContext is context judgment infrastructure for AI agents, RAG systems, and retrieval workflows. Instead of only retrieving more sources, it evaluates caller-provided candidate context for freshness, provenance, confidence, utility, and source profile, then returns decision-ready output: cite, verify, refresh, background, watch, or exclude. The current front door is evaluate_context, with demos, trust gates, and reference adapters included.
