Vexon is the AI agent built into Semavex. It answers questions about your live databases in plain language, builds the semantic model that captures your business definitions, and turns answers into reports it can email, schedule and edit on request.
Each agent has one job, and each works from your semantic model, so every answer uses your business definitions.
Describe your business in your own words. Vexon reviews every table, writes the business context, describes each column with a confidence level, infers relationships, flags sensitive data and builds your measures, all for you to review before the model goes live.
Ask a question, or start from the questions Vexon suggests from your own model. It answers from your live data, keeps the context for follow-up questions, flags what deserves a closer look and explains the why behind a number in plain language.
Save any answer as a report, then just ask: email it as a PDF, send it every Monday, remove a metric. Vexon sets up the schedule, explains what a change affects and applies it once you approve.
Type a question in plain language, in any language, or pick one of the questions Vexon suggests from your model.
Vexon turns the question into a structured query plan (metrics, dimensions, filters, joins and date range) using your semantic model and your database's index metadata.
Semavex checks the plan against your semantic model and a whitelist of allowed operations, translates it into parameterized SQL for your database engine and runs it read-only.
The result comes back as a formatted report with a plain-language interpretation. Save it, export it to CSV or PDF, or hand it to the reporting agent.
Language models are good at understanding a question, not at being exact every time. So Vexon's only output is a typed query plan, never a query string. Deterministic code in Semavex turns that plan into SQL, which makes every query predictable, read-only and traceable to the plan that produced it.
Vexon cannot change your data, platform settings or user permissions. Scheduled reports rerun their saved plan without calling Vexon again.
{
"metrics": ["Gross profit"],
"dimensions": ["Casino", "Quarter"],
"dateRange": "Last 4 complete quarters",
"sort": "Quarter, ascending"
}Read-only access, with database credentials encrypted (AES-256-GCM).
5:29Ask a question in plain language, get a report your team can trust, then email it, schedule it or edit it just by asking. Follow a finance and operations team through seven real workflows, from suggested questions and a custom P&L to the Reporting Agent, a year-over-year expense analysis, a cash-flow paradox and slot-machine rankings.
5:29 · 7 chapters
3:54From raw database to business-ready semantic model in minutes. We connect a SQL Server operations database and a PostgreSQL finance database, then Vexon turns raw tables into business meaning: context, column descriptions with confidence levels, relationships, measures and synonyms, all built from a plain-language business goal.
3:54 · 4 chaptersVexon is the AI agent built into Semavex. It answers questions about your live databases in plain language, builds the semantic model that captures your business definitions, and turns answers into reports it can email, schedule and edit on request.
No. Vexon is part of Semavex and is included in both plans, Professional and Enterprise. You use it inside the Semavex app, and its AI usage draws on your Semavex AI credit wallet.
No. Vexon produces a structured query plan. Semavex validates that plan and translates it into parameterized SQL with deterministic code, then runs it read-only. The model never writes or runs a query itself.
PostgreSQL, SQL Server, MySQL and MariaDB. Snowflake support is coming soon.
Not a general one. You talk to Vexon in plain language and it keeps the context of the conversation, but it only works on your data: every answer comes from a query plan checked against your semantic model, and it does not answer general questions or improvise.
No. Report content is not used to train any AI model, and data read from your database is discarded once the report is delivered.
Vexon is built on Claude, by Anthropic, and constrained to produce structured query plans rather than free text or SQL.
You can write questions and business descriptions in any language. The Semavex app and website are available in English and French.
Book a demo and see how Vexon answers questions in the language of your business.