stop writing
SQL from scratch — let AI understand your schema and generate SQL for you.
Oracle SQLcl
25.2+ introduced one of the most important developer features in years: native
MCP (Model Context Protocol) server support. With this, Oracle becomes the
first major database platform to offer a standardized way for AI assistants
like ChatGPT and Claude to securely understand your schema, generate SQL, and
even run queries — all from plain English.
In this post,
we’ll walk through:
- What MCP is and why it matters
- How SQLcl can act as an MCP Server
- Setting up VS Code to use natural language chat
with your database
- Real demos — from simple queries to advanced
analytics
- How AI can generate reports and dashboards from
your Oracle data
Let’s get
started.
What is MCP
(Model Context Protocol)?
MCP is an
open-source standard for connecting AI models to real systems.
Just like USB-C
gives your devices a universal connector, MCP gives AI a universal connector to
databases, filesystems, tools, and workflows.
With MCP:
- An AI assistant can inspect your database schema
- Generate SQL queries
- Execute them only with your permission
- Produce dashboards, summaries, or reports
- Plug into developer tools like VS Code
Oracle SQLcl
now supports this protocol natively — no separate servers, plugins, or bridges
needed.
SQLcl as an
MCP Server
SQLcl 25.2 introduced the new command:
sql -mcp
Running this
transforms SQLcl into a live MCP Server, allowing AI clients to talk to the
database through a safe, controlled interface.
Example:
C:\Users\Rajeshwaran
Jeyabal> sql -mcp
---------- MCP SERVER STARTUP
----------
MCP Server started
successfully on Fri Nov 28 19:27:09 IST 2025
Press Ctrl+C to stop the
server
----------------------------------------
This server
acts as a secure gateway — the AI can propose SQL, but you must approve
execution.
Prerequisites
Before
beginning, ensure the following:
- Visual Studio Code v1.101 or later
- Oracle SQL Developer Extension for VS Code
(version 25.2.0+)
- At least one Oracle Database connection with
password saved
Once this setup
is complete, the chat panel in VS Code will automatically detect the MCP tools
exposed by SQLcl.
Saving a
Database Connection for MCP
First create a
saved connection in SQLcl — MCP relies on this.
C:\Trash> sql /nolog
SQLcl: Release 25.3
Production on Thu Nov 27 20:46:15 2025
idle> conn
hr/hr@localhost:1521/ora23ai
Connected.
hr@ORA23AI> connect
-savepwd -save hr_local_mcp
Name: hr_local_mcp
Connect String: localhost:1521/ora23ai
User: hr
Password: ******
Now the MCP
server will be able to use this connection automatically.
See It in
Action — Talking to Your Database in Natural Language
Open the Chat
Panel in VS Code.
You will notice
SQLcl’s MCP tools automatically appear on the left sidebar.
This confirms
that VS Code + AI Assistant + SQLcl MCP Server are connected.
Example 1 —
“Show me all tables with their data volume”
You simply ask
the AI:
Can you
connect to the HR database and show me all tables with their data volume?
Claude /
ChatGPT will:
- Check available MCP tools
- Propose a SQL query
- Ask for your confirmation
- Execute it through SQLcl
- Return a neat table of row counts
This ensures
safety — nothing runs without approval.
Claude always asks permission prior
to executing any queries. We can see exactly what it plans to execute before
approving.
Example 2 —
Analyzing Employees Hired in 2008
You ask:
Show me
employees hired in 2008, grouped by department, with average salary for each
group. Also show the highest-paid employee in each department.
AI generates a
multi-part SQL including: - Filtering by hire year
- GROUP BY analyses
- JOINs to departments
- Window functions to identify the highest salary
And after you
approve, the results come back instantly.
Example 3 —
AI-Generated Analytics Dashboard (Markdown Report)
Now let’s ask
something more advanced:
Can you
create a comprehensive HR dashboard showing: department headcount, salary
ranges, recent hiring trends, and identify any departments that might need
attention based on salary distribution or hiring patterns. Create the report as
MarkDown
The AI will:
- Pull multiple SQL queries
- Aggregate the results
- Identify anomalies/trends
Generate a
clean Markdown report that looks like this:
HR Dashboard
(Generated by AI)
1. Department
Headcount
Sales — 34 employees
IT — 19 employees
Marketing — 12 employees
Finance — 11 employees
2. Salary
Distribution
IT: Highest variance, indicates
mixed seniority
Sales: Several outliers with
high commission-based compensation
Finance: Narrow salary band →
stable structure
3. Recent
Hiring Trends
Hiring spikes in Q2 and Q4
Marketing hired heavily last
year due to campaign cycle
4. Departments
Needing Attention
IT — high salary spread
Marketing — rapid hiring,
potential onboarding load
This entire
dashboard was produced using:
- Your HR schema
- SQL generated automatically
- Data analyzed via AI
- Result formatted into Markdown
No SQL written
manually.
Why This
Matters
This workflow
fundamentally changes how developers work with Oracle:
- No need to remember every syntax detail
- No need to hand-craft queries
- Faster exploration of schema and data
- Insightful reports without BI tools
- Safe execution (AI asks before running SQL)
- Works locally — no cloud dependency
With SQLcl MCP
Server, your database becomes conversational.
Conclusion
Oracle has
taken a major step forward by bringing MCP support directly into SQLcl.
Combined with VS Code AI chat, developers can now:
- Query databases in plain English
- Let AI generate complex joins and analytics
- Approve execution safely
- Produce dashboards and reports instantly
This is the
future of database development — not replacing SQL, but removing the friction
around writing it.
