What Is Pinecone Nexus?

Pinecone, the vector-database company, launched Nexus in public preview in July 2026 and made it generally available in August 2026. Nexus is a “knowledge engine” for AI agents: it ingests enterprise content, has subject-matter experts curate it into domain-specific knowledge, and compiles that knowledge ahead of time so agents can retrieve it in a single call through KnowQL, a declarative query language. It adds field-level access control, citations, confidence scores, and lineage, runs in the customer's cloud, and supports customer-chosen models. Pinecone reports over 90% lower token costs than agentic RAG and the top score on Sierra's τ-Knowledge customer-support benchmark.

The Core Difference

Nexus is built for documents. It chunks, curates, and compiles content into a knowledge store, then answers from that curated copy. Tursio is built for structured data. It builds metadata — a context graph of joins, measures, dimensions, and business aliases — and uses it to query the original database directly, returning live results. One answers “what does our policy say?”; the other answers “what is our delinquency rate by branch this month?”

Where Nexus Falls Short for Structured Data

• Answers come from a compiled copy. Knowledge is compiled ahead of time, so answers are only as fresh as the last compile. Balances, transactions, and claims change by the minute; a pre-compiled view goes stale.
• Chunks are not tables. Document chunks and curated passages cannot reliably produce exact aggregations, filters, and joins across millions of rows. Precise numbers require executing queries against the database.
• Curation is ongoing human work. Nexus relies on subject-matter experts to shape knowledge. For structured estates with thousands of tables (one Tursio customer has 18,000+), hand curation does not scale.
• Another copy of sensitive data. Compiling knowledge means moving and storing derived data in a new system, which must be secured, governed, and kept in sync with the source.

How Tursio Is Different

• Metadata, not data copies. Tursio's context graph stores lightweight logical models of the database — far smaller than the data — so nothing needs to be re-ingested when values change.
• Live queries against the source. Each question compiles to native SQL and runs on the system of record. Answers are always current and exact.
• Automated, not curated. The context graph is inferred from schema, query history, and BI definitions and tuned from feedback — no expert curation pipeline.
• On-prem, governed, flat-priced. Single Docker image in your perimeter, Entra ID and RBAC, BYO LLM, flat license.

Better Together

Enterprises need both kinds of answers. A Nexus deployment can cover policies, procedures, and support content while Tursio covers live operational and financial data — both exposed to the same agents over MCP. Where the question needs a number from a database, the agent should call Tursio.

Side-by-Side Comparison

Bottom Line

Pinecone Nexus is a strong answer for making document knowledge agent-ready. It is not built to answer precise, live questions from structured systems of record. Tursio builds the metadata and queries the source directly, so every agent gets current, exact, governed answers — without copying the data.