--- title: Nexyron: Behavioral Analytics & Graph Intelligence on Your Mac url: https://nexyron.tech/ description: Graph database, behavioral analytics, churn prediction and causal decisioning in one engine that runs on your Mac. A local-first alternative to enterprise ontology and decision platforms. --- # Nexyron: Intelligence that learns how your business behaves *Business behavioural intelligence, by Athlera.* One engine turns operational data into behaviour you can predict, explain and act on. It runs on your Mac. Nexyron is in private pilot. Access is arranged through https://nexyron.tech/request. Documentation: https://nexyron.tech/docs (Markdown index at https://nexyron.tech/docs.md). The mathematics behind the decision layer: https://nexyron.tech/method. ## Six systems, or one A connected question normally needs a graph database, a vector store, a semantic layer, a feature store, a time series system and somewhere to run models. Each one is a separate contract, a separate copy of the data, and a separate place for the answer to go wrong. The joins between them become your problem. The systems a connected question normally needs: - Graph database - Vector store - Semantic layer - Feature store - Time series - Decision runtime Nexyron is one query surface, one transaction path, one storage layer. The graph is the mechanism, not a component. ## How it works: five moves, in one place Every behavioural analytics product does one or two of these and hands you off for the rest. The value is in the sequence staying intact. ### 1. Observe: your records go in as they are Spreadsheets, exports, mailboxes, database dumps and edge lists are previewed, planned into a graph shape with temporal streams and field mappings, then written through the same canonical path as every other change. A bad sheet fails as one item, not as the whole run. - csv, parquet, xlsx, json, jsonl, eml, mbox, sql, arrow, tsv, txt - plus exchange formats for data that is already connected ### 2. Abstract: it names the concepts it finds A subject is not a row. Nexyron reads the records and the written company knowledge and registers business concepts as contracts, each with a query back to the evidence that produced it. What was measured, what was extracted from a document and what was inferred stay separable. - business concepts registered as contracts, evidence attached - features, snapshots and datasets derived from what was observed ### 3. Lens: a question becomes reusable A Lens binds a business concept to an analytical purpose and becomes executable. It selects its own indicators under a stated intent, runs in the background against the current data, writes a report with its evidence attached, and can be run again next month without being rebuilt. - churn, forecast, utilisation, delay root cause, anomaly, journey - scored, explained and calibrated, with uncertainty stated ### 4. Decide: then estimate what an action would change Correlation names a group. It does not tell you whether acting on that group is worth anything. Nexyron carries interventions, policies and scenarios as first-class objects, estimates uplift rather than association, and refuses to be confident where the treated and untreated populations do not overlap. - uplift and overlap checked before a number is believed - policies compared against each other before anyone acts ### 5. Remember: every answer leaves something behind Reports, knowledge assertions, features, model outputs, Lenses and recorded decisions stay addressable and searchable. The second question starts from what the first one built, which is the only version of this that compounds rather than resets. - one search across records, documents, concepts and reports - saved reports, knowledge books, recorded decisions and their outcomes ## Questions that need the whole picture Each of these fails in a warehouse because the answer lives in three places at once: what the subject did, where it sits in the organisation, and how long it has been that way. That is one behavioural query here. ### Industrial fleet and field service **Question:** Which of these assets fails before the next service window, and does moving the window actually prevent it? Sensor and work-order history becomes a temporal stream per asset. The graph carries the site, the route, the part and the technician. A Lens scores failure risk; a policy estimates what a rescheduled visit is worth against the cost of making it. What the answer draws on: history per asset, network position, payoff estimate. ### Subscription and recurring revenue **Question:** Which accounts are behaving the way last year’s departures behaved, and has any intervention ever changed that? Behaviour is the payment stream, the support history, the product events and the elapsed time since each, read together. Recorded interventions and their outcomes turn the second question from an opinion into an estimate grounded in what actually happened. What the answer draws on: behaviour cohorts, intervention history, what actually worked. ### Manufacturing and supply **Question:** This order slipped four days. Which supplier, route and prior order share the cause? Delay is a path problem before it is a statistics problem. The engine traces routes and shared bottlenecks across the order network in the same query as the numbers, so the explanation and the total come from one place. What the answer draws on: path analysis, community detection, root cause report. ### Insurance and financial crime **Question:** Are these fourteen claims independent, or one network with three shared intermediaries? The question is structural, and structure is what the engine is for. The engine’s network analysis exposes the shared actors, and meaning-based search over the claim documents finds the connections the identifiers do not show. What the answer draws on: shared intermediaries, network structure, document similarity. ### Multi-site operations **Question:** Is this site underperforming, or is it the cohort that site happens to serve? The org tree is a graph, so a roll-up from site to district to division is a traversal rather than a join. Comparing like with like means holding the cohort constant, and the engine does that matching itself. What the answer draws on: org roll-up, like-for-like cohorts, fair comparison. ## The Lens is the unit, not the dashboard Most tools organise around a dashboard or a chat window. A dashboard freezes one shape of one question. A chat window forgets. A Lens is neither: it is a written contract between a business concept and an analytical purpose, and it is the thing that gets reused. It states its subject and its intent, chooses its own indicators from what the data can actually support, declares the buckets it needs and refuses to run when the population is too thin to mean anything. When it finishes it leaves a report, a model artifact and a set of features that the next Lens can build on. The anatomy of a Lens: - **Subject**: the registered abstraction the Lens reasons about - **Intent**: the governing objective, in plain language - **Indicators**: four to ten complementary features, selected under that intent - **Feasibility**: exact training, selection and test populations, with minimums - **Evidence**: the observed series and aggregates behind every claim - **Output**: a report, a model artifact and reusable features ## The engine: what is actually in it Competitors in this category describe outcomes and keep the mechanism private. The mechanism is the argument here, so it is on the page. - **63 graph algorithms**: centrality, community, flow, paths, similarity, structure - **42 causal and decision procedures**: uplift, propensity, doubly robust, off-policy, bandits - **20 model procedures**: scoring, calibration, ROC, gain, uncertainty, intervals - **6 data styles, one query surface**: graph, temporal, semantic, abstraction, model, decision ### Time is a property, not a table Streams are declared on the owner and carry indexed paths, rollups and wall-clock buckets. History, rollup, bucket aggregate and grouped bucket aggregate are procedures over the same graph, so a temporal answer and a structural answer are one query. ### One semantic index, four kinds of thing Node embeddings, edge embeddings, abstractions and abstraction artifacts share a single approximate-nearest-neighbour index and return one row shape. Retrieval adds bounded graph expansion, source references, score components, provenance and temporal hints. ### Embeddings computed locally Local ONNX and Core ML embedding providers run in-process. The macOS release converts the default model to Core ML at build time and ships the package as an application resource, so an installed app needs no Python and no external inference call. ### Cypher-first, and honest about it Nexyron is not a Neo4j drop-in and does not pretend to be. Syntax carried from Neo4j, APOC, GDS or SQL fails deliberately, and every failure returns a diagnosis, the closest real symbols, the pages to read and a runnable example. ### Written like a database, because it is one Committed work survives a crash or a power cut. Working sets larger than memory still run. Analysis is built to use the machine you already own rather than demand a bigger one, and when one machine is no longer enough the same engine grows across several and keeps answering when one of them fails. ### Reachable from where you work HTTP, WebSocket and Bolt for the server. Embedded in your own process through Rust, Python, Node, C and WebAssembly. Connected to Claude, Codex and ChatGPT for the agent. Every route reaches the same connected model, so the answer is the same answer. ## Deployment: your data does not need to go anywhere ### The Mac app: it runs on the machine in front of you Open the app and the full analytical engine is running on your own computer. Your records, your documents and every answer built from them stay on hardware you own. Nothing is uploaded to be understood, nothing waits on someone else’s cloud, and the confidentiality of your business data is a property of the machine, not a promise in a vendor’s contract. - Installed like any app: no data platform project, no integration team - AI assistance works with your data on this machine only, never across the internet - Privacy and compliance reviews get short, because the data never leaves the building ### Your servers: clustered on infrastructure you control When one machine is no longer enough, the same engine grows into a resilient cluster inside your own data centre, run and governed by your own team. ### Operated by Athlera: a dedicated instance we run for you Single-tenant, not a shared analytics cloud. The same build, the same interface, the same data model, with the operations handled. ## Agents: the assistant you already use can hold it Nexyron plugs straight into Claude, Codex and ChatGPT desktop. Your assistant reaches your business memory, your files, your saved reports and the behavioural analytics engine as tools it already knows how to use. No new interface to learn, and every answer comes back in a reliable, structured shape. Access is asked for, never assumed. Choosing which business to open, writing changes and taking in new knowledge each require your confirmation, in the assistant you are talking to. And the connection works with the assistant your team runs today, without an upgrade project first. What the connection carries that a plain chat assistant does not: - **Remembers your business**: facts, files and decisions persist across every conversation, and compound - **Absorbs whole files**: spreadsheets, mailboxes and documents become part of the model, not pasted text - **Computes, never estimates**: cohorts, forecasts and network analysis run in the engine, so the numbers are exact - **Turns questions into Lenses**: an analysis worth keeping becomes reusable, and runs again on live data - **Keeps working after the chat**: long reports run in the background and come back finished - **Remembers what you decided**: actions and their outcomes are recorded, so next quarter you can ask what worked ## How Nexyron compares: the short version The nearest thing to Nexyron is the ontology platform: a category built on the same conviction, that a model of how the business behaves has to exist before an answer can be trusted. We are building it from the other end. ### The model is derived, not commissioned Every governed system in this category depends on a semantic model authored in advance: an ontology, a modelling layer, a semantic view, a curated knowledge store. That authorship is the real cost, and it is paid before the first answer. Nexyron derives its abstractions from the records and the documents, then lets an operator correct them. ### The decision layer is in the database Uplift, propensity overlap, doubly-robust estimation, off-policy comparison and causal optimisation are procedures over the same graph the data lives in, called from the same query. Not a notebook next to the warehouse. Not a chart with a recommendation written underneath it by a language model. ### The deployment is an application, not a programme The incumbent arrives as an implementation with a timeline and a team. Nexyron arrives as an app that starts a database, ingests your files and answers a question the same afternoon, on hardware you already own, with the option to become a cluster later. ## Where the product is today Straight answers about the current state, so you know exactly what you are evaluating. - **In private pilot**: Nexyron is running with design partners today. Names, numbers and case studies appear here when partners agree to stand behind them publicly, and not before. - **Mac app first, server anywhere**: The desktop application is built for the Mac. The server and cluster run on whatever infrastructure you choose. - **Built for behavioural insight**: The focus is how your customers, assets and operations behave over time, and why. Cohorts, churn, forecasts and the effect of interventions, not another dashboard of totals. - **Local by design, more local soon**: The understanding of your data already runs on your machine. Running the assistant’s reasoning locally as well is the next step on the roadmap. ## Industries Each one has a mature toolchain that answers the question it was built for, and stops at the same boundary. These are the questions on the other side of it. ### Fraud and financial crime Structure hidden inside records you already hold. - [Insurance fraud](https://nexyron.tech/industries/insurance-fraud): A score judges one claim at a time. Rings are built so that every claim survives that test. - [Financial crime and AML](https://nexyron.tech/industries/financial-crime-aml): Everyone suspects which rule has never produced a filing. Nobody can prove it, so it stays. - [Forensic accounting](https://nexyron.tech/industries/forensic-accounting): An intercompany account that reconciles perfectly can hide the fact that nothing was ever transferred. ### Integrity and assurance Behaviour that only has a shape over time. - [Procurement integrity](https://nexyron.tech/industries/procurement-integrity): Across eleven tenders, in rotation. No screen looks there, because nobody joins award data to contract data. - [Internal audit](https://nexyron.tech/industries/internal-audit): It passes every test you run, because every test you run is a test about the entry. - [Insider risk](https://nexyron.tech/industries/insider-risk): Your identity platform knows who can reach the data. Your behaviour platform knows what they did. ### Operations and capital Your own record as the evidence of your own fragility. - [Asset reliability](https://nexyron.tech/industries/asset-reliability): Your prediction platform is trained on a field your technicians do not fill in. - [Supply chain risk](https://nexyron.tech/industries/supply-chain-risk): Your risk platform can tell you a supplier is risky. It has never seen one of your orders. - [Private equity diligence](https://nexyron.tech/industries/private-equity-diligence): Retention improved after they invested in it. Neither fact establishes that the spend caused it. ## Bring one question and your own data A working session on your machine with your files. If the answer is not in the data, Nexyron is built to say so rather than produce a confident number. Request access: https://nexyron.tech/request *Nexyron is the engine. Athlera Intelligence is the interface.*