Oracle ships a built-in MCP server inside SQLcl (25.2+) that lets AI models query your Oracle database through the Model Context Protocol. Ask questions in plain English — the model writes the SQL, runs it through a named SQLcl connection, and explains the result. Everything runs locally, so no data leaves your machine.
Bring your own
Claude Sonnet 4.5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
A named SQLcl connection (username, password/wallet and TNS or connect string). Use a read-only database user scoped to the schemas the agent needs. Data stays local via SQLcl.
How models use it and what it is built for.
The SQLcl MCP server exposes Oracle's command-line client as a set of MCP tools. Models can list the saved connections, run SQL and PL/SQL against a chosen connection, inspect schema objects (tables, views, indexes, constraints), and read data-dictionary metadata. Because it runs through SQLcl on your machine, it uses your existing named connections and wallet — no connection string is pasted into the agent, and Oracle states no data is sent to Oracle services. A common read-only alternative wraps an Oracle JDBC connection for schema, query and metadata access. It is the go-to for chatting with enterprise Oracle databases without hand-writing SQL.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
List the tables in the HR schema and tell me which ones look like fact tables.
Retrieve all employees hired in the last 30 days from the HR schema.
Show me the indexes on the ORDERS table and flag any that look redundant.
Summarise the columns and data types of the CUSTOMERS view and flag any PII.
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Oracle Database
Claude Sonnet 4.5
Claude Sonnet 4.5 writes correct Oracle SQL and PL/SQL and reasons over schema metadata reliably. Use Haiku 4.5 for simple read queries to keep costs low.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
It is Oracle's built-in MCP server, shipped inside SQLcl 25.2 and later. It lets AI models query and inspect an Oracle database through natural language by translating requests into SQL and running them through a named SQLcl connection on your machine.
No. The SQLcl MCP server runs locally and executes SQL against your database through your existing named connections. Oracle states that no data is sent to Oracle services — data sovereignty is maintained on your machine.
You need SQLcl 25.2 or later, which includes the built-in MCP server. It connects to your Oracle Database using named connections or a wallet, and works with any MCP-capable client such as Claude Desktop, Cline or MCP Agent Studio.
Use a dedicated database user with read-only privileges scoped to only the schemas the agent needs, and never connect as SYS or SYSTEM. For exploration, point the agent at a non-production database or a read replica during setup.
Both let AI models query a relational database in natural language. Oracle MCP runs through SQLcl and speaks Oracle SQL and PL/SQL against named connections, while Postgres MCP connects via a standard connection string and generates PostgreSQL SQL. Choose based on which database you run.