VantegrateNative API integrationIntegrate Databricks with AI agentsthat query your lakehouse in plain language
Vantegrate's AI agents query your lakehouse through Databricks SQL warehouses: Metrix turns a business question into a query on your governed tables, with read-only permissions managed by Unity Catalog, and sends the answer back on WhatsApp. AI comes to your data: your tables keep living in your Databricks platform.
How the data flows
What does it mean to integrate Databricks with Vantegrate's AI agents?
Integrating Databricks with Vantegrate's AI agents means that Metrix queries your lakehouse to answer business questions, instead of an analyst writing every notebook or query by hand. Unlike a fixed decision-tree bot, which only returns what was preloaded, the agent turns the question into a SQL query on your real data, with read-only permissions governed by Unity Catalog.
The connection runs through Databricks SQL warehouses, and the exact technical path is validated case by case before we start. Our principle applies: AI comes to your data, not your data to the AI. Your tables stay in your Databricks platform: the agent queries and returns the answer, it doesn't copy or move the database.
What do you want to connect to your Databricks?
Pick what you need and we'll write the message for you.
1 · Pick what to solve
2 · What our team receives
Vantegrate
online
What the agent queries in your Databricks
The agent works on your lakehouse read-only: it turns the question into a SQL query on the governed tables you authorized and returns the answer where your team already works.
Queries on your governed tables
Metrix turns the business question into a SQL query on the lakehouse tables you authorized, and runs it on a SQL warehouse with the Unity Catalog context.
Read-only access
The agent comes in with USE CATALOG, USE SCHEMA and SELECT on the tables it needs, plus CAN USE on the SQL warehouse, and no write or admin permissions.
Answers where the team asks
The agent returns the answer in plain language on the channel where your team already works, such as WhatsApp. The data is queried in your platform, not copied somewhere else.
How the integration works, step by step
Vantegrate handles the setup and the rollout. Your team authorizes, validates and starts using it.
Discovery and technical path
We map the workspace and the Unity Catalog catalogs and schemas to query, and we define and validate the path (SQL warehouse, OAuth or personal access token) case by case before connecting.
Connection through SQL warehouses
We connect through a SQL warehouse with the official JDBC/ODBC driver. With OAuth machine-to-machine and a service principal, each access token is valid for one hour and renews automatically.
Minimum permissions in Unity Catalog
We create or use a service principal with read-only permissions: USE CATALOG, USE SCHEMA and SELECT on the tables it needs, plus CAN USE on the SQL warehouse. With a PAT, we use tokens limited to the API scopes it needs.
Business context and validation
We load the business context so the agent understands your tables and metrics, we test with real questions and we roll out in phases.
Querying the lakehouse through SQL warehouses vs. asking an analyst for the report
Not every way of getting data out of the lakehouse is the same. Asking an analyst for every query leaves gaps: the question becomes a ticket, the answer arrives when the data team has a free slot and, in the meantime, the decision waits. Querying through SQL warehouses with governed access turns the question into SQL and runs it on the spot, without that delay.
The difference shows in every question: the agent queries the governed tables with read-only permissions and answers instantly, not on the data team's next turn. And because access goes through Unity Catalog, every query stays limited to what you authorized and is logged, instead of getting lost among exports and stray spreadsheets.
| Dimension | Asking an analyst for the query | Metrix through SQL warehouses |
|---|---|---|
| Access | Notebook or manual export | Read-only SQL warehouse |
| Permissions | Ad hoc, sometimes too broad | USE CATALOG, USE SCHEMA and SELECT in Unity Catalog |
| Authentication | Shared credentials | OAuth with a service principal, or a scoped PAT |
| Response time | Waits for the data team's turn | On the spot, in plain language |
| Traceability | Scattered across exports | Every query is logged |
Comparison of integration patterns; the exact scope is defined with your team based on your workspace and your Unity Catalog setup.
Your data stays in your platform: the agent queries the lakehouse through SQL warehouses with read-only permissions and returns the answer; it doesn't copy or move your tables. AI comes to your data, not your data to the AI.
You've seen how it connects. Want to walk through your case?
Tell us how your Databricks is set up and we'll tell you what data we need and where the agent connects.
An agent connected to your Databricks vs. a standalone tool
| Standalone AI tool | Agent connected to your Databricks | |
|---|---|---|
| Data source | A copy that drifts | Your source of truth, live |
| Where your data lives | In a third-party system | In your environment, encrypted end to end |
| Result of each question | A one-off export | An answer from live data, where your team works |
| Permissions and audit | Broad and hard to audit | Read-only role and a log of every query |
| Reaching production | Often stalls as a pilot | Validated, phased rollout |
Why add an agent that queries your lakehouse
Third-party figures, each with its published source, to size up the platform and the context. None of the figures is a Vantegrate result.
5x
Faster queries for Databricks customers compared with three years ago, according to the company itself
95%
Of generative AI pilots never reach production with measurable business impact
The speed figure is a claim Databricks publishes about its own platform, not a Vantegrate result. The exact connection path to your Databricks is validated case by case before we promise it.
Which agents run on your Databricks
The same read-only connection through SQL warehouses feeds the agents in the suite: they query the governed lakehouse and act on what it returns, without taking your database.
Metrix
Data agent: answers business questions in plain language by querying the governed lakehouse tables through SQL warehouses, with the Unity Catalog context, without anyone writing the notebook.
Explore Metrix
Sellium
Sales agent on WhatsApp: uses the business context that lives in your lakehouse, such as segments or history, to serve customers and quote with up-to-date data, without leaving the chat.
Explore Sellium
Revio
Outbound marketing agent: acts on the signals that already live in your lakehouse, such as a score or a segment, to send the WhatsApp reactivation message at the right moment. A strong fit for customers in markets where WhatsApp is the default channel.
Explore RevioThe integration doesn't take your data anywhere
The principle is simple: AI comes to your data, not your data to the AI. The agent queries your lakehouse through SQL warehouses, with the read-only permissions you authorize, and the agents run on certified cloud infrastructure.
- Connection through SQL warehouses with the official JDBC/ODBC driver; OAuth machine-to-machine authentication with a service principal (recommended) or a scoped personal access token. No scrapers and no manual exports.
- Read-only, governed by Unity Catalog: USE CATALOG, USE SCHEMA and SELECT on the tables it needs, plus CAN USE on the SQL warehouse. No write or admin permissions.
- You operate and certify your Databricks workspace and your Unity Catalog; the agent queries the lakehouse, it doesn't take the database.
- The agents run on Salesforce or Oracle Cloud Infrastructure, whose SOC 2 Type II and ISO 27001 certifications belong to those platforms, not to Vantegrate or your Databricks account, with traceability for every query.
Frequently asked questions about the Databricks integration
What data and IT teams usually ask before connecting an AI agent to their Databricks lakehouse.
What permissions does the agent need in Databricks?
What permissions does the agent need in Databricks?
Read-only access through Unity Catalog: USE CATALOG, USE SCHEMA and SELECT on the tables it needs, plus CAN USE on the SQL warehouse. No write or admin permissions. The agent queries the governed tables you authorized and returns the answer in plain language; Vantegrate handles the setup and the rollout, and your team authorizes the permissions, validates and starts using it. Meet the data agent at Metrix.
Where does my data stay?
Where does my data stay?
In your Databricks platform. The agent queries the lakehouse through SQL warehouses and returns the answer; it doesn't copy or move your tables anywhere else. Access uses minimum read-only permissions governed by Unity Catalog, and every query is logged. The principle is that AI goes to your data, not your data to the AI.
OAuth or a personal access token?
OAuth or a personal access token?
The recommended option is OAuth machine-to-machine with a service principal, for security and automatic renewal: each access token is valid for one hour and renews on its own. You can also use a scoped personal access token, restricted to the scopes it needs. The final choice is validated during discovery, based on your workspace's policies.
Can I combine it with Snowflake or BigQuery?
Can I combine it with Snowflake or BigQuery?
Yes. Metrix can query several analytics environments: each one connects with its own read-only setup. You can have the Databricks lakehouse governed by Unity Catalog and, alongside it, other data warehouses, and the agent answers with each one's context according to the permissions you authorized.
Metrix, the data agent on WhatsApp
How the agent answers business questions by querying your governed tables, without anyone building the report.
Learn moreAll integrations
The full map: which CRMs, ERPs, payment gateways, data warehouses and tools the AI agents connect to.
Learn moreAI agents for enterprises
The map of the Vantegrate suite: sales, marketing, documents, data and logistics.
Learn moreLet's connect AI agents to your Databricks
Tell us which business questions you'd like to answer from your lakehouse and we'll show you how Metrix queries it through SQL warehouses, which read-only permissions we request in Unity Catalog and how the answer reaches you on WhatsApp. A 30-minute conversation, no commitment.
Francisco Morales, co-founder, takes your call. We reply on WhatsApp within 4 business hours, no strings attached.
