Scattered knowledge
Procedures, reports, presentations, messages, and business documents remain spread across many spaces.
Document and AI infrastructure
Make your company knowledge usable by your teams, applications, and AI agents.
Nexus collects, structures, and governs your documents to retrieve useful information, produce source-linked answers, and power AI use cases — in an environment your company controls.
Self-hostableAPI-firstGoverned accessBYOK
AI does not lack answers. It lacks governed context.
General-purpose assistants can produce text. They do not inherently know your company’s current documents, access scopes, or the sources an answer should rely on.
Files handled repeatedly
Users have to upload files, copy passages, or rebuild context for every new request.
Access that is hard to control
Giving an assistant access must never mean opening all company knowledge.
Answers without a clear origin
Without retrievable documentary references, it is difficult to review, validate, or deepen an answer.
Isolated applications
Each new application risks rebuilding its own storage, context, and rules.
Opaque usage and costs
Without shared observability, the models used, latency, errors, and AI spending remain hard to manage.
Full product demonstration · 20 minutes · English
See Nexus in action
Understand Nexus through a complete, step-by-step walkthrough of the product and its interface. Allow approximately 20 minutes for the full demonstration.
Complete walkthrough · English audio
One shared foundation, from source to use.
Nexus turns scattered documentary sources into structured, authorized, reusable knowledge without imposing the same interface on every use case.
Sources
DocumentsURLsEmailNotionSlackJiraOther applications
Nexus
Structure
Authorize
Retrieve
Evidence
Uses
Teams
Applications
AI agents
Same corpora
Same permissions
Same references
Same observability
Application sources are available according to integrations and the Nexus ecosystem configuration, including through Composio.
The interface or model can change. Nexus retains the document structure, permissions, references, and operating rules.
Simple conversational access to authorized knowledge
Nexus WebUI is Nexus’s conversational chatbot. Users ask questions in natural language and receive an answer alongside the passages and sources retrieved within their scope.
- Questions and multi-turn conversations
- Synchronous or streaming answers
- Selection of an authorized scope
- Choice among authorized models
- Citations, sources, and user feedback
Service → Workspace → Project
Selecting a scope makes search easier, but never creates an entitlement. Permissions enforced by Nexus remain authoritative.
Simple use cases without shifting complexity to users
Nexus lets teams start with very concrete questions, then progressively extend the same foundation to other teams and applications.
Find a procedure
Nexus searches only authorized documents and returns the procedure with its sources.
Understand a visual report
With the Heavy profile, Nexus can use text, tables, and extracted images, then return the relevant passage, page, and image when available.
Verify information
Agentic Search can cross-check documents, reread relevant passages, and distinguish verified, partial, or not-found results.
Build a business application on Nexus
A company can use Nexus’s API-first foundation to create its own products, portals, assistants, or workflows without rebuilding the document layer, permissions, and observability.
From raw documents to searchable knowledge
Nexus keeps source files in object storage controlled by the company, then orchestrates extraction, normalization, chunking, and indexing.
Import → Extract → Structure → Index
PDFDOCXPPTXXLSXMarkdownTXTHTMLCSVWebVTTAsciiDoc
The formats actually accepted depend on active ingestion profiles. TXT is supported through configuration.
Enhanced processing for visual documents
The Heavy profile is designed for documents with many images, charts, tables, or pages that are difficult to extract. It can use OCR, table analysis, image description, and page-level information retention.
Visual document
- Images
- Charts
- Tables
- Complex pages
Heavy processing
- OCR
- Table analysis
- Image description
- Page detection
Structured artifacts
- Text
- Passage
- Page
- Image
- Metadata
Usable answer
- Retrieved passage
- Relevant page
- Available image
- Associated analysis
During a search, Nexus can therefore return the passage, page, relevant image, and its analysis when those elements are available in ingestion artifacts.
Two levels of search, depending on the question
Classic document search answers most questions quickly. Agentic Search investigates requests that require several searches, rereading, and more structured verification.
Document search
Question → Relevant passages → Context → Answer → Citations. Suitable for everyday search, procedures, summaries, and direct documentary questions.
Agentic Search
Complex question → Requirements to check → Successive searches → Open documents → Gather evidence → Verify → Answer.
Termination date identified
Verified
Same notice period in every contract
Partially verified
Automatic penalty provided for
Not found in authorized sources
The verification status results from checks orchestrated by Nexus. The model does not decide alone that an answer is verified.
Knowledge remains governed in Nexus — and therefore within the environment controlled by your company.
The documentary scope is defined by Service, workspace, and project. The calling application is identified separately for authentication, quotas, caching, traceability, and observability: it does not define the corpus and cannot broaden rights.
Tenant → Service → Workspace → Project
Server-side permissions
Rights are applied by Nexus during search and again when documents are opened. A model cannot decide to broaden its scope.
Self-hosting
Nexus can be deployed on infrastructure controlled by the company: locally, on a server, a VPS, or suitable cloud infrastructure.
BYOK and model choice
Provider keys are encrypted and never exposed in the browser. The company can use its own contracts, choose providers, and track the related costs.
Observability
Nexus tracks queries, models, providers, latency, costs, tokens, caches, citations, feedback, ingestion, and component states.
- BYOK gives a company more control and optimization levers for AI costs. It is not an automatic savings guarantee.
- Self-hostable does not mean every model necessarily runs locally. Nexus can use external providers configured by the company.
The assistant and model can change. Corpora, permissions, sources, evidence, and usage rules remain governed in Nexus, inside the environment your company controls.
One foundation for teams, applications, and agents
Nexus does not confine company knowledge to a single chatbot. Its capabilities are available through the WebUI, API, and three specialized MCP servers.
WebUI
Nexus
API · MCP servers
nexus-api · User OAuth
Lets an assistant query corpora, documents, and RAG functions accessible to the signed-in user. It exposes neither administration nor ingestion.
nexus-doc-kg · Service token
Lets authorized agents explore documents, resources, relationships, neighbors, subgraphs, and context packs in the document Knowledge Graph. Read-only access.
nexus-code-kg · Service token
Gives developer agents a structural memory of applications. Most tools are read-only; a controlled regeneration requires a specific scope and explicit confirmation.
API-first
The API lets applications reuse ingestion, search, RAG answers, citations, documents, scopes, media, projections, graphs, and authorized administration functions.
Model Context Protocol
MCP means Model Context Protocol. These servers let a compatible assistant or agent use Nexus capabilities as structured tools, with the permissions and scopes enforced by Nexus.
Nexus already provides the foundation for Nexus Pocket, Nexus Knowledge Studio, Silio, and Aurora Atlas. The same API-first architecture lets each company create its own applications, products, and agents.
How is Nexus different from ChatGPT, Claude, Gemini, or another online AI assistant?
These assistants are excellent general-purpose interfaces. Nexus serves a different need: making company knowledge durably available, governed, and reusable across several AI use cases.
Online assistants
general-purpose interface and capabilities
knowledge dependent on the model and supplied context
spaces, files, and connectors according to the chosen offering
experience mainly centered on the assistant
uploads and copy-paste when context must be rebuilt
Nexus
self-hostable document infrastructure
corpora organized by Service, workspace, and project
server-side permissions
citations and documentary references
API for applications
MCP for authorized agents
choice of providers and models
usage, cost, and performance tracking
Nexus does not aim to replace every assistant. When they are connected and authorized, it gives them structured, governed, retrievable company knowledge.
For teams that also need to know how it works
The first level explains value. These details provide the architecture, operations, and integration evidence technical teams need.
Frequently asked questions about Nexus
Which use case should you make reliable first?
A procedure to retrieve, reports to query, an internal assistant to connect, or a business application to build: start with the concrete need and the relevant documentary scope.
Discuss my use case