Prepare data, add business context, share it, and see how it's used.
Updates run automatically, so your team and AI agents always work with the latest data.
Publish, add business context, ask from an AI agent, and see every access.
The problem isn't the AI. It's the data you give it. Without trusted data and business context, AI misreads your data, repeats the same work, and wastes tokens.
AI misreads your data and your business rules.
AI cleans and prepares the same data every time someone asks.
AI reads far more data than the question needs.
Better AI answers don't start with a better prompt. They start with better-prepared, better-understood data.
The AI adds up every order. The West's 13.7% return rate inflates its sales, so it looks like the leader.
The AI follows the approved definition, which matches how the business counts net sales.
Net sales = sum ofsales
where
return_flag = "N"One example from the demo sales data. Business context helps AI answer more reliably; it doesn't guarantee a correct answer.
Monitor feeds back into Add Context: what people ask tells you what context to add next.
Clean, join and calculate with recorded steps in Exploratory.
Then schedule a refresh: Exploratory re-queries the source and reruns every step, so the data is always ready.
Every dataset has the same views: Summary for each column's distribution and missing values, Table to filter and aggregate, and Wrangling Steps to see exactly how each number was made.
You don't have to take anyone's word for it.
In the Metadata view, Generate / improve with AI drafts the business context from the data and from the recorded wrangling steps. You review it, fix what's off, and approve it.
The Access Log shows who used the data, when, and how, and the exact query they ran.
The queries and their run times give you hints on how to improve the business context.
And if you see access you didn't expect, you can shut off a channel such as MCP or the CLI.
Where your data lives, who can use which data, and how it was used. With the Data Catalog, you decide all of it and can check all of it.
Publish to the Local Data Catalog and the data never goes to Exploratory Server.
Turn MCP and CLI access on or off for each dataset, at any time. In Exploratory Desktop, you can also turn MCP and CLI off for the whole app.
On Server, an AI app signs in as the person, so it sees only the data that person may see. Everywhere, AI agents get only approved business context.
On the Data Catalog page, the dataset owner can see in the Access Log who used which data, when, through which channel, and the exact query. Each dataset has its own Access Log tab too.
Data that shouldn't leave your company stays in the Local Data Catalog on your PC; data your team needs goes on the server. The main features work in both.
| Local Data Catalog | Server Data Catalog | |
|---|---|---|
| Where the data lives | Your PC | Exploratory Server |
| Who can use it | You, in Exploratory Desktop | People in your organization, with their own permissions |
| Sharing | None | Team sharing and a web page for each dataset |
| Scheduled refresh | Yes | Yes |
| Summary, Table, Wrangling Steps | Yes | Yes |
| Metadata (business context) | Yes | Yes |
| Access Log | Yes: Exploratory Desktop, MCP, CLI | Yes, for the dataset owner: web page, MCP, CLI, API, with IP address |
| MCP and CLI on or off per dataset | Yes | Yes |
| MCP and CLI on or off for the whole app | Yes, in System Settings | — |
| AI apps through MCP | Desktop MCP: Claude Code, Codex, Antigravity, Cursor | Server MCP: Claude, ChatGPT, Microsoft Copilot, Gemini, Claude Code. The AI signs in as the user and sees only what that user may see. |
| Best for | Personal work, and companies whose policy keeps data off external servers | Shared, governed data for teams and organizational AI |
Some companies don't allow data on external servers, or AI features inside other apps, but do allow a company-approved AI client such as Claude Code.
With the Local Data Catalog, you prepare data in Exploratory, keep it on your PC, and query it from the AI tool your company already approved. You decide which datasets that tool can reach, and the Access Log records each query.

Turn on MCP and install the plugin for Claude Code, Codex, Antigravity or Cursor in one click. Turn the CLI on or off in the same place.
From any project in Exploratory Desktop, pick a dataset from the Local or Server Data Catalog, build a query visually (select columns, filter, aggregate, join), and get the result.
Import the data straight into Exploratory Desktop for analysis, or check and visualize it in a web browser and download the rows you need as a CSV to use in Excel and other tools.
Find the dataset they need, read its approved business context, and run a query through MCP. They can reach only the data that the account they signed in with through OAuth has permission for.
Import a source, clean it with recorded steps, and schedule a refresh.
Generate and approve the business context, then query the data yourself or connect your AI app through MCP.
Don't have an account? Start a 30-day free trial
One place in Exploratory to prepare, verify, and share trusted data that people and AI agents can understand and query efficiently. Every dataset has the same views: Summary, Table, Wrangling Steps, Metadata and Access Log. It runs on your PC (Local) or on Exploratory Server for your team.
No. Scheduled refresh keeps your data up to date, and Summary, Table and Wrangling Steps let you verify it. In Exploratory Desktop, you can build a query in the UI and import just the data you need into any project. AI agents are one more way to use it.
No. Publish to the Local Data Catalog and the data stays on your PC; it never goes to Exploratory Server.
Not by itself. The dataset stays on your PC, but the query results your AI client receives go to that AI provider, under your company's agreement with them.
You decide which datasets your AI client can reach, and the Access Log records each query.
Yes. Turn MCP and CLI access on or off for each dataset, at any time, on Local and on Server. In Exploratory Desktop, you can also turn MCP and CLI off for the whole app in System Settings.
Open the Access Log on the Data Catalog page, or the Access Log tab of a dataset. It shows who used which data, when, through which channel, and the exact query. Only the dataset's owner sees it.
Business context tells the AI what the data means: what one row is, what each column means, how metrics are defined, and how to aggregate.
It helps AI answer more reliably; it doesn't guarantee a correct answer. AI agents get only the context you approved.
On Server, people use each dataset with their own permissions, and an AI app signs in as the person, so it sees only what that person may see. On both Local and Server, you can turn MCP and CLI access on or off per dataset.
The Access Log records each access. On Server: the web page, MCP, the CLI and the API, with the IP address. On Local: Exploratory Desktop, MCP and the CLI. Only the dataset's owner sees it.
Yes. Schedule a refresh: Exploratory re-queries the source and reruns every recorded step, so the dataset in the Data Catalog stays up to date, on Local and on Server.
Prepare your data once. Give people and AI agents the context and access they need.