MARE started with a simple question: do we really need to embed everything?
Instead of chunking and embedding the full dataset, MARE uses a small semantic navigation index and lets an agent investigate data through MongoDB tools.
In a 10K-incident benchmark, it used just 604 vectors vs 60,000 for RAG, while scoring 19/20 vs 18/20 on held-out questions.
RAG remains faster for simple lookup. MARE explores where agentic retrieval helps on multi-hop, investigative questions.
ACGC (Agent Context Garbage Collector) is a lightweight sidecar for agentic AI systems. It dynamically selects, compresses, and retrieves relevant context within a fixed token budget using a memory DAG, semantic recall, and archival retrieval. Evaluations show 46–82% lower token usage while maintaining or improving quality—making agents cheaper, faster, and more memory-efficient without major application changes.
