Pinecone Nexus
Knowledge layer deployable in your cloud providing governed, faster, more accurate, and lower-cost knowledge for agents.
Pinecone Nexus is a knowledge layer now generally available that runs in your cloud (AWS, Google Cloud, or Azure) so documents and compiled knowledge never leave your infrastructure. Nexus lets you supply model credentials and run inference on the models you choose (including open-weight), supports ensembles per workflow, and compiles a governed knowledge layer you can download as an archive. Agents, chatbots, AI search, and recommendation systems query Nexus via a single interface (KnowQL). Nexus sits on Pinecone Database as its retrieval foundation and integrates with Pinecone Marketplace. In τ-Knowledge benchmark runs (as of Aug 4, 2026) adding a Nexus knowledge layer reduced tool and model calls about in half, lowered cost per task (example: a $1.45 task became $0.53), and improved or retained accuracy (GPT-5.2 saw 12% more accuracy and ~80% cost reduction; GPT-5.5 held accuracy at substantial cost reduction). In an internal test on a customer support queue, the agent's autonomous resolution rate rose from 25% to 55%.