Skip to main content
You point Nexus at your sources and curate them into a context. Agents then query that context with a single call and get back a structured, grounded answer with citations, instead of re-assembling context from raw chunks on every request. Traditional RAG hands an agent ranked chunks and leaves it to search, stitch, and re-search. Nexus does that work once, upstream, so agents spend their budget on reasoning instead of re-orienting.

Quickstart

Prepare your first context, then query it.

Key concepts

Contexts, manifests, artifacts, sessions, and the Query API.

How Nexus works

Curated artifacts for retrieval, then answers composed over a retrieval SDK.

Bring your own cloud

Deploy Nexus in your own cloud account.

​
What you can do

  • Turn raw sources into queryable knowledge. Upload documents or connect a data source, then curate them into a searchable context. There is no chunking pipeline or eval set to hand-build.
  • Ask in one call. Agents issue a KnowQL query (ask plus scope, optionally a typed shape) and get back a grounded, cited answer. One query is a turn. Multi-turn conversations are sessions.
  • Compose across domains. A single query can span multiple contexts, so knowledge is a graph of contexts rather than one monolithic index.
  • Keep answers grounded and scoped. Answers carry citations, and access is scoped to your Pinecone project.

​
Why Nexus

The most expensive part of an agent is not reasoning. It is knowledge acquisition. Agents built on raw retrieval burn tokens re-orienting on every call and loop through retrieve, evaluate, and re-retrieve cycles. Nexus compiles the knowledge once and serves it on every call:
  • Cheaper than agentic RAG. The expensive work happens once at curation time, not per query, so agents spend far fewer tokens.
  • Faster than a retrieval loop. A single query replaces the multi-step retrieve, evaluate, and re-retrieve cycle.
  • Trusted, grounded answers. Each one carries citations and is scoped to your Pinecone project, running on the Pinecone Database underneath.

​
What Nexus is not

  • Not a vector database. The Pinecone Database sits underneath Nexus as the retrieval substrate. Nexus is the engine layer above it.
  • Not RAG. RAG returns ranked chunks for a human to read. Nexus compiles typed, grounded artifacts purpose-built for agents and returns answers.
  • Not an agent framework. Nexus is knowledge infrastructure: query in, answer out. Your agent decides what to do with the answer.