Your data already has the answer. Getting it shouldn’t take weeks.
Nexatron connects to the systems your teams already run: Salesforce, NetSuite, your ERP, your warehouse. It works out what the data means on its own, then answers questions in plain English. No migration project before the first answer. No waiting in the analytics queue.
Our partner ecosystem
AI can read your data. It can’t know what you mean by “revenue”.
The definitions and join logic that make a number trustworthy live in people's heads and old SQL, not in your warehouse. Without them, AI answers on raw tables and gets confidently wrong.
Nexatron works out what your data means.
It reads your existing schemas and queries to build the metric definitions, joins, and rules your AI needs. As the business changes, Nexatron proposes updates and routes them to a person for review, so every change stays versioned and governed.
One question, answered from the sources you already run.
Enterprise data rarely lives in one platform. Nexatron connects to your warehouses and business apps where they are, with no warehouse migration. Joining several sources in one answer is on by default for new workspaces, and a workspace admin can turn it off.
Snowflake
Google BigQuery
Databricks
Amazon Redshift
Salesforce
NetSuite
HubSpot
ServiceNow
Amazon S3 / Parquet
DuckDB
Microsoft Excel
Google SheetsUse Claude with your governed business data.
Connect Claude to your business through Nexatron
Bring supported business applications and databases into Nexatron, then use its MCP tools from Claude. Start with the sources and permissions your organization has approved.
1. Connect your sources
Salesforce, NetSuite and supported warehouses supply the data. Check the connector catalog for availability and scope.
2. Define the business meaning
Nexatron brings source context, metric definitions and configured access controls into the analytical workflow.
3. Ask through Claude
Claude calls the analytical tools exposed by the Nexatron MCP server. Available tools and data depend on workspace configuration.
Claude is a trademark of Anthropic, PBC. Nexatron and dotSolved Inc are not affiliated with, sponsored by or endorsed by Anthropic.
Three ways to ask. One engine checks every answer.
Conversational BI
Ask in chat and get an answer you can check, with the lineage behind every figure.
AI-powered dashboards
Dashboards built on certified metrics. When a number changes, the lineage shows why.
Analytics agents
Agents that query your data under the asker's permissions, with every call logged and auditable.
A fixed suite of golden questions runs against every release.
Welcome back, Sam
How it works
Three steps, and none of them are a data project
The usual way to get answers across several systems is to move everything into one place first, agree what every field means, and build reports on top. That is months of work before anyone asks a question. This skips it.
Connect the systems you already run
Salesforce, NetSuite, your ERP, your data warehouse. Nothing moves and nothing is copied. Your data stays where it is, under the same permissions it has today.
Most teams are answering questions the same week they connect.
It works out what your data means
Instead of a modeling project, Nexatron reads each system and maps out what it holds: the customers, the invoices, the opportunities, and how they relate. Your analysts review and approve the definitions that matter, instead of building all of them from scratch.
This is the part that normally takes a quarter and a consultancy.
Your team asks
"How much did we invoice this quarter, by region?" Typed, spoken, or asked from the AI assistant your team already uses. The answer comes back in seconds, with the working shown, so it can be checked before anyone acts on it.
No ticket, no queue, no waiting for someone technical to be free.
Why you can trust the number
It tells you when it doesn’t know
The expensive mistake is not a tool that says “I can’t answer that”. It is one that answers anyway, confidently, and nobody notices until a decision has been made on it. These are real messages, copied straight from the product. Try them in your own demo and check the wording matches.
- You ask
Break our accounts down by account type
It answersSalesforce does not expose RecordTypeId on the requested object to this connected user. I did not drop the record-type dimension or substitute a different breakdown.
The easy failure is quietly answering a slightly different question and not saying so.
- You ask
Which customers with open deals are late paying us?
It answersThe certified entity link can multiply records at this aggregation grain, so we stopped rather than double-count.
Matching these up the obvious way double-counts, and the total looks perfectly reasonable. You see this where a workspace admin requires certified links between sources.
- You ask
Compare pipeline in the CRM with invoiced sales in the ERP
It answersThis question combines data from sources that aren't linked by a certified entity link yet. Ask your workspace admin to declare and certify the link between these connections.
Matching customers by name is a guess. A guess that is usually right is the dangerous kind. You see this where a workspace admin requires certified links between sources.
- You ask
How many opportunities are named ACME-999?
It answersI checked Salesforce but nothing matched. That is a count of zero rather than a confirmed business total: it means nothing matched the filters this query applied.
"None" and "nothing matched your filters" mean different things. People have acted on the wrong one.
- You ask
Show me every contact's social security number
It answersI won't query or expose passwords, authentication tokens, government identifiers, or bank-account details. No Salesforce query was executed.
It stops before it looks, rather than looking and then hiding what it found.
One agreed definition. Everyone gets the same answer.
People asking in chat. Agents calling over MCP. Dashboards subscribing to a metric. All of them hit the same governed semantic layer and get the same number.
Switch between Analyst and Agent: the question and the answer match, down to the same confidence score.
Enterprise expansion ARR accelerated +22%. New pricing lifted mid-market ACV by $14K. The partner channel closed its first two enterprise accounts in week 11.
Your product. Our answer engine.
Drop governed conversational analytics into your own app. Your customers ask under your brand and get answers that respect your security model, with lineage attached.
Any MCP agent becomes a governed analyst.
Connect Claude, Copilot, or your own agents to the Nexatron MCP server. They get the same governed context and confidence score your people get. Agent queries run under row-level security and land in the same audit log as human queries. With bring-your-own LLM keys, model calls run on your provider contract.
EMEA and LATAM. EMEA closed at 112% of plan on enterprise expansion; LATAM hit 106% in the first full quarter of the partner program.
AI assistants play by the same rules as your people.
Every query runs under row-level security and masking, whether a person or an agent sent it. All of it lands in the audit log.
Flexible deployment
Governance
Security
Frequently asked questions
What Nexatron is, how it stays accurate, and how it keeps your data governed.
What is Nexatron?
Nexatron is an enterprise AI analytics platform by dotSolved. You connect your data warehouse. Then a person in chat or an AI agent over MCP can ask it questions and get answers governed by your security rules. A governed answer comes back with the exact query that produced it, its data lineage, and a confidence score; a question it cannot answer under your rules is refused or clarified instead.
What is a semantic layer?
A semantic layer is the shared definition of what your data means: which metrics are certified, how tables join, and which business rules apply. Nexatron's semantic layer, MetricQL, captures those definitions once so that every question is answered against trusted, governed logic instead of raw tables. It is what lets AI answer correctly instead of guessing at column names.
How does governed natural-language-to-SQL work, and how is it kept accurate?
Nexatron translates a question into SQL against your certified metrics and semantic layer, not against raw tables. The model's output is treated as untrusted and validated against the schema and your governance rules before any query is allowed to run. Each answer returns the rows, the exact generated SQL, its lineage, and a confidence score, so accuracy is verifiable rather than assumed.
Which AI assistants can connect over MCP?
Any MCP-capable assistant, including Claude, ChatGPT and Cursor, can connect to Nexatron's hosted MCP endpoint. The connection uses Streamable HTTP with JSON-RPC 2.0 and OAuth 2.1 with PKCE and Dynamic Client Registration. Agents can run governed asks and discover and execute certified metrics, and, where a workspace enables the optional action platform, take actions with dry-run or execute and run multi-step workflows. All of it runs under the same governance and audit trail as a human user.
Which data sources can I connect?
Nexatron ships nine warehouse and database connectors: Snowflake, Google BigQuery, Amazon Redshift, Databricks, PostgreSQL, MySQL, Microsoft SQL Server, Amazon S3/Parquet, and DuckDB. DuckDB is also the embedded engine behind cross-source federation: when one question needs two or more of your connections, Nexatron queries each one and joins the results inside your tenant. Federation is on by default for new workspaces, and a workspace admin can turn it off. Beyond the warehouses, a separate and growing library of business-app connectors (Salesforce, QuickBooks, NetSuite, ServiceNow, HubSpot, Stripe, Microsoft Dynamics 365, Atlassian and more) connects SaaS applications with per-user OAuth, so reads run under each user's own permissions. The full live catalog is at nexatron.io/connectors.
How is my data isolated and secured in a multi-tenant platform?
Every query filters by tenant, and four-layer row-level security is enforced on every request whether it comes from a person or an agent. Your warehouse credentials are stored as AES-256-GCM encrypted values in the connections table, managed through the Connections UI. Your data and queries are never used to train foundation models, and Nexatron is SOC 2 Type I attained (Type II in progress), ISO 27001 certified, and GDPR-aligned.
What comes back with an answer?
Every governed answer ships with what you need to check it: the result rows, the exact query Nexatron generated, the data lineage behind it, and a confidence score. Instead of an unexplained number, you get one you can audit and defend. A low confidence score tells people and agents to verify before acting.
Do people and AI agents get the same answers?
Yes. Whether a question arrives from a person in chat or from an AI agent over MCP, it runs through the same semantic layer, the same certified metrics, and the same four-layer row-level security. Every query, from a person or an agent, lands in the same audit log. There is no separate, less-governed path for machines.