NEXORA · KNOWLEDGE-BASE ENGINE
Your documents, answering with receipts.
Nexora is Plainsight's knowledge-base engine: it makes your documents searchable in plain language and answers with cited sources, showing each person only what their permissions already allow.
In plain terms: Nexora is not another chat assistant. It's the engine underneath one. Your assistant, or any application you run, sends it a question; it comes back with the answer, the exact source, and nothing the asker isn't allowed to see.
An engine, not an assistant
You don't open Nexora. Your tools do. It's the component that finds the answer, wherever the question gets asked. Three ways to plug it in:
Built into our AI Assistant
Choose the assistant and Nexora is already inside: every answer your people get is grounded, cited and permission-aware, with nothing to integrate. One package, one price. Nexora is included, not an add-on.
Meet the AI Assistant →Plugs into what you already run
An API your intranet, portal or internal tools can call. Same cited answers, same permission model, inside the applications your people already use.
Feeds every MCP tool via the Brain
As the document layer of the Plainsight Brain, assistants like Claude and ChatGPT can search it too. One engine, every tool.
How the Brain works →Why we built it
Every “chat with your documents” demo looks the same: point an indexer at a folder, ask a question, applause. Then it meets a real company. Five versions of the same contract. Scanned pages. A colleague asking in Dutch about a document written in English. And the question the demo never asks: is this person even allowed to see that file?
We hit all of that while building the document layer of our own Brain, and the tools we tried kept failing on the same points. Answers without sources. Permissions as an afterthought. Confident nonsense when the honest answer was “it's not in there”.
So we built the engine we needed. It reads documents properly before indexing them, cites the paragraph it answers from, enforces who may see what on every single chunk, and says so when it doesn't know. Boring virtues. They're what make people trust the answers.
What Nexora does
The searching itself is a discipline. These are the four parts that separate a knowledge base people trust from a folder with a chat window.
Ask in your own language
Ask the way you'd ask a colleague. Search understands meaning as well as exact wording, and a question in one language finds the answer in a document written in another.
Answers with receipts
Every answer cites the paragraph it came from, so people can check instead of trust. When the answer is not in the documents, Nexora says so rather than serving the closest noise.
Per document, not per promise
Row-level security down to document and chunk. The engine never returns content someone could not already open in the source system.
Documents read properly
Tables in contracts, scanned pages, presentations where the point is in a picture. Findable means read well first, and duplicates get collapsed on the way in.
Built to survive scale
We run Nexora on our own nine knowledge bases before we run it on yours. Two lessons from that logbook, kept on purpose.
Silent beats leaky
When two layers of the stack disagreed about who could see what, Nexora answered with nothing rather than leaking. It looked broken. It was doing its job. That behaviour is designed in, not luck.
Fast under load
Typical search answers in about 13 milliseconds. Early on, heavy ingestion pushed some queries to 40 seconds, so we re-engineered until search stays fast while the indexer eats a document backlog. Your team feels that difference by Tuesday.
Frequently asked questions
Is Nexora another chat assistant?
No. Nexora has no chat window of its own. It's the retrieval engine other tools call: our AI Assistant uses it as its knowledge layer, and your own applications can query it through an API. If you want the chat experience, that's the AI Assistant. If you want the engine inside something you already have, that's Nexora.
How is Nexora different from pointing an indexer at a folder?
The folder trick demos well and degrades fast. Nexora reads documents properly at ingestion (tables, scans, slides), collapses duplicate versions, cites the exact paragraph in its answers, and enforces permissions per document. And when the answer is not in the documents, it says so instead of guessing.
How are permissions enforced?
Per document and per chunk, with row-level security in the index itself. Each person only retrieves what their permissions already allow in the source system. If layers of the stack ever disagree about access, Nexora returns nothing rather than leaking.
Which document types does it handle well?
The messy real ones: contracts with tables, scanned pages, slide decks where the point is in a picture, and the five slightly different versions of the same file. Findability starts at ingestion, not at query time.
Which languages does it support?
Ask in the language you think in. Search understands meaning as well as exact wording, so a question in one language finds answers in documents written in another. We tune every deployment on the languages your company actually uses.
What happens when the answer is not in the documents?
Nexora says so. An engine that always answers is an engine you cannot trust; the honest “it's not in there” is a feature, and it took real engineering to build.
Can we run Nexora without the AI Assistant?
Yes. Nexora is the retrieval engine and works with any front end that speaks to it, including your own applications. The AI Assistant is the ready-made chat front end we offer on top, and the Plainsight Brain is the wider memory layer both plug into.
What does running it cost?
With the AI Assistant: nothing extra. Nexora is part of the platform, included in the one price. Standalone, it runs on open components in your own environment, so you pay your own infrastructure at cost instead of a per-query search bill. The honest total depends on document volume and usage; we size it with you before anything is committed.
Want answers with receipts from your own documents?
Bring one messy document set, we'll show you what cited, permission-aware search looks like on it.
Contact an expert