Pinecone Systems Inc., an artificial intelligence infrastructure company providing fully managed vector databases, Wednesday launched the public preview of Pinecone Nexus, which curates and distributes enterprise knowledge for AI agents.

The era of agentic AI is here and it’s built on top of data. How that data is delivered to AI agents is becoming the pipeline that decides the difference between accurate answers and forgetfulness. The industry continues to work hard on how to take raw information, package it and deliver it in a way that AI can use.

Employees already readily handle data from multiple sources by knowing where to look, thanks to an internal compass rooted in institutional memory. Ask them a question and they recollect that they need to look in wikis, human resources docs, meeting notes, support tickets and financial records to pull out what they need. AI agents don’t come with this level of knowledge. Instead, they require frameworks, graphs and systems of knowledge to rebuild this sort of sense of “I don’t know this off the top of my head, but I know where to find it” every time you call them.

Pinecone Nexus is designed to curate access to knowledge for AI agents, allowing them to reason across dozens of files at once, without grabbing small chunks.

What Nexus brings to the industry

AI agents need access to data, so Nexus first provides connectors to retrieve and import it. This includes local file upload, Box and Microsoft OneLake live today. In the pipeline coming soon are Google Drive, Slack, GitHub, Notion, Confluence and S3.

The engineering team setting up an AI agent works within a project container called a Workspace that organizes data into Contexts, which define data as knowledge sets or domains. Individual teams build their own workspace and use that to maintain their own cognitive knowledge base for their AI agents.

This flows through Nexus’ curation layer, which uses templates called Manifests to turn raw documents into knowledge artifacts that match their contents.

Essentially, Nexus provides a framework that explains the underlying “where to find what you’re looking for” for the AI agent. This is similar to how an employee with over three years at the company knows which part of the financial archives to consult for acquisition data or where to look for the most recent business logic for the web portal.

That’s different from prompt engineering, where users or engineers must teach the agent where to look at query time, whereas Manifests allow the agent to understand where knowledge lives during its curation. That brings domain experts into the loop as information is categorized, ensuring it’s ready when the AI agent reaches for it.

In benchmarks, Pinecone said Nexus demonstrated high performance and accuracy.

“We can easily stand up a vector database and run RAG (and agentic search) over our documentation corpus,” Jesse Barbour, chief data scientist of Q2 Holdings Inc., an Austin-based financial technology solutions company. “The hard part is getting an agent to reliably and efficiently assemble the right knowledge for genuinely difficult questions.”

According to Barbour, Nexus answered complex support questions with 95% accuracy. It also kept token costs low, making it an enticing knowledge layer as AI inference prices rise.

In another example, an unnamed data protection and security vendor tested Nexus on a tranche of documents with widespread information, including municipal meeting minutes, meaning that questions could cross multiple documents and chunks.

Curating 598 documents into 12 structured artifact types cost $2.31 and took 34 minutes. Subsequent queries achieved about 90% accuracy, compared with a 65% baseline for the industry-standard retrieval-augmented generation pipeline.

Image: Shutterstock/Nepool