Summary
On July 18, 2026, Pinecone made Nexus generally available: a knowledge engine that curates scattered enterprise data into structured, queryable contexts that AI agents access through a query language called KnowQL, shifting token spend to a one-time curation step instead of repeated per-query retrieval.
What changed
Pinecone released Nexus (GA) with Workspaces, Contexts, Manifests, and the KnowQL query language, plus connectors for local files, Box, and Microsoft OneLake (Google Drive, Slack, GitHub, Notion, Confluence, and S3 planned). Standard and BYOC (bring your own cloud) deployments are offered.
Why it matters
Nexus reframes retrieval for agents: instead of re-embedding and re-searching on every query, teams curate a structured knowledge layer once. Pinecone reports roughly 9-15x lower token cost in legal use cases alongside higher accuracy and task completion, which matters as agent workloads turn per-query retrieval into a real operating expense.
Evidence excerpt
Nexus shifts token spend to a one-time curation step, reducing costs by approximately 9-15x in legal applications while improving accuracy and agent task completion rates.