Oracle Vector DB MCP. Run complex vector searches and inspect schema from chat.
Gemini Works with every AI agent you already use
…and any MCP-compatible client
Just plug in your AI agents and start using Vinkius.
Oracle Vector DB MCP Server lets your agent run vector similarity searches directly on Oracle Database 23ai. You can execute native VECTOR_DISTANCE queries, inspect table structures with `describe_table`, manage HNSW/IVF indexes via `list_vector_indexes`, and query tables using standard SQL.
It brings enterprise-grade database metadata management into natural conversation.
What your AI agents can do
Describe table
This tool lists table columns and their explicit data types, including any VECTOR fields defined in the schema.
Execute sql query
It runs arbitrary SQL queries against the Oracle runtime via ORDS. Remember to limit rows fetched for stability.
Get database version
This tool fetches and reports the exact version banner of the running Oracle DB Runtime.
You execute native Oracle 23ai VECTOR_DISTANCE queries, finding the most similar records using metrics like cosine.
The tool lists all accessible tables and describes their columns, specifically noting which ones are defined as VECTOR types.
You run any standard SQL query against the Oracle database for data analysis that isn't vector-based.
The agent lists all specialized HNSW and IVF vector indexes attached to your tables, helping you plan searches.
You retrieve row counts, optimizer statistics via table_stats, and verify the Oracle runtime version using get_database_version.
Ask AI about this MCP
Supported MCP Clients
OAuth 2.0 Compatible Gemini Waiting for input…
Oracle Vector DB: 7 Tools for Database Operations
Use these seven tools to describe tables, execute SQL, list indices, or run complex vector similarity searches against your Oracle database.
Make your AI actually useful.
Add this MCP to Claude, Cursor, or Windsurf and your AI stops guessing. It gets real tools to look things up, take action, and handle the stuff you keep doing by hand.
Start using Oracle Vector DB on Vinkius019d75ebdescribe table
This tool lists table columns and their explicit data types, including any VECTOR fields defined in the schema.
019d75ebexecute sql query
It runs arbitrary SQL queries against the Oracle runtime via ORDS. Remember to limit rows fetched for stability.
019d75ebget database version
This tool fetches and reports the exact version banner of the running Oracle DB Runtime.
019d75eblist tables
It returns a list of all accessible tables within your current Oracle schema.
019d75eblist vector indexes
This tool lists the specialized AI Vector search indexes (HNSW, IVF) that have been instantiated across your database tables.
019d75ebtable stats
It retrieves critical table cardinality and optimizer statistics for capacity planning or performance tuning.
019d75ebvector search
The tool executes a vector similarity search using Oracle 23ai's native VECTOR_DISTANCE function on specified columns.
Choose How to Get Started
Build a custom MCP for your own tools, or connect a ready-made integration from our catalog.
Build Your Own
Turn any API into an MCP. Import a spec, define Agent Skills, or deploy with MCPFusion.
- Import from OpenAPI, Swagger, or YAML specs
- Create Agent Skills with progressive disclosure
- Deploy to edge with MCPFusion framework
- Built in DLP, auth, and compliance on every call
- Real time usage dashboard and cost metering
- Publish to catalog or keep private
Make Your AI Do More
Start with Oracle Vector DB, then connect any of our 4,800+ other servers whenever your AI needs more. One click, no limits.
- Use this MCP plus 4,800+ others, all in one place
- Add new capabilities to your AI anytime you want
- Every connection is secured and compliant automatically
- Track usage and costs across all your servers
- Works with Claude, ChatGPT, Cursor, and more
- New servers added to the catalog every week
Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by Oracle Database. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.
VINKIUS INFRASTRUCTURE
Cloud Hosted
Managed infra
V8 Isolated
Sandboxed per request
Zero-Trust Proxy
No stored credentials
DLP Enforced
Policy on every call
GDPR Compliant
EU data residency
Token Compression
~60% cost reduction
Works with Claude, ChatGPT, Cursor, and more
The Model Context Protocol standardizes how applications expose capabilities to LLMs. Instead of operating in isolation, your AI gains direct access to external platforms, live data, and real-world actions through secure, standardized connections.
This server provides 7 capabilities that interface natively with Claude, ChatGPT, Cursor, and any MCP client. No middleware. No custom integration required.
Database metadata should never require jumping between three different panes just to check a column type.
Today, if you want to know what data types are available or confirm if a column is set up for embedding storage, you usually have to run `DESCRIBE` in one window, then switch over to another tool just to list the tables, and maybe check a third dashboard to see the row count. It's a messy workflow.
With this MCP server, you tell your agent: 'Show me all vector-ready columns.' The agent runs `list_tables` and then uses `describe_table`, giving you a single, clean list with everything you need in one chat window.
Oracle Vector DB MCP Server helps you perform complex searches using the `vector_search` tool.
Manual vector search used to mean scripting out a query that calculated distances, making sure it handled cosine vs. Euclidean metrics, and then sorting by proximity—a multi-step process prone to error.
Now you just ask: 'Find the top 5 documents most similar to this embedding.' The agent runs `vector_search` natively inside Oracle DB, delivering ranked results without any custom code from you.
What you can do with this MCP connector
Your agent runs vector similarity searches straight out of Oracle Database 23ai. You don't need to leave your chat client; you just talk to it. This server lets your AI client execute native VECTOR_DISTANCE queries, check the whole database structure, and run custom SQL—all without writing complex JDBC calls.
Vector Search Operations
You use the vector_search tool when you need to find similar records across your data. You specify columns that hold vector embeddings and define a distance metric like cosine or Euclidean. The tool executes Oracle's native VECTOR_DISTANCE function, returning the nearest neighbors based on how close their vectors are. For general data analysis that isn't about similarity, you run standard SQL using execute_sql_query.
Remember to keep your row limits low when running these queries for stability.
Schema and Structure Management
You can inspect exactly what's in the database with a few simple calls. To see every table accessible in your schema, just use list_tables. If you need details on a specific table—like figuring out which columns hold embeddings or what data type they are—you run describe_table. This tool lists all column names and their explicit data types, making sure to flag any fields defined as VECTOR types.
To plan your vector searches, the agent checks specialized indexes using list_vector_indexes; this shows you every HNSW and IVF index attached to your tables. You can also get a general overview of how much data's in play by checking the table stats with table_stats, which pulls critical cardinality counts and optimizer statistics for capacity planning.
System Health Checks
Before you run big queries, you gotta check if the system is running right. You can confirm the exact compatibility level of your setup using get_database_version, which reports the live Oracle DB Runtime banner. For a complete picture of what's available to query, these tools give you direct access: list tables (list_tables), describe table schema (describe_table), retrieve vector indexes (list_vector_indexes), run similarity searches (vector_search), execute arbitrary SQL (execute_sql_query), pull performance data (table_stats), and verify the environment version (get_database_version).
019d75eb-9634-7229-8a1b-e0cd1fe251a5 
How Oracle Vector DB MCP Works
- 1 First, you subscribe to the server and provide your connection details (ORDS URL, credentials).
- 2 Next, you ask your AI client a question—like 'Find documents similar to this embedding.'
- 3 The agent automatically translates that request into tool calls (
vector_searchorexecute_sql_query) and returns the results directly.
The bottom line is: it turns complex, multi-step database operations into simple conversations with your AI client.
Who Is Oracle Vector DB MCP For?
This server is for data professionals who spend too much time context-switching between a chatbot and a SQL query window. If you're an ML engineer building RAG pipelines or a DBA needing to verify vector index integrity, this saves hours of manual scripting.
You test embedding queries on Oracle 23ai's native vector engine during the development of Retrieval-Augmented Generation (RAG) pipelines.
You inspect vector indexes, review table stats, and check schema configurations through natural conversation instead of writing complex DESCRIBE scripts.
You run ad-hoc SQL queries or perform similarity searches against production data without needing to open a dedicated client like SQL Developer.
What Changes When You Connect
- The
vector_searchtool executes native Oracle 23ai VECTOR_DISTANCE queries. You don't have to switch tools or write complex PL/SQL just to find the most similar records; you talk to your agent, and it handles the math. - You gain full visibility into your database structure by using
describe_table. This tool shows all column types, letting you quickly pinpoint which fields are structured as VECTORs for embedding storage without guessing. - Forget running basic lookups in a separate UI. With
execute_sql_query, the agent lets you run any arbitrary SQL query directly from your chat window, making ad-hoc data analysis faster than ever. - The
list_vector_indexestool is crucial for maintenance. It lists every HNSW and IVF vector index across your tables, giving DBAs a clear inventory of where their specialized AI search capabilities live. - You get performance insight instantly with the
table_statstool. Need to know if a table can handle more data? Runtable_stats. It delivers row counts and optimizer metrics immediately.
Real-World Use Cases
Finding Missing Documentation
A support agent needs to find documentation similar to 'API Rate Limits Overview'. Instead of writing a complex SQL JOIN, they ask their agent: 'Find the top 5 documents related to API rate limits.' The agent runs vector_search using cosine distance, returning the exact document titles and distances.
Schema Validation for New Data
An ML engineer is setting up a new data source. They first run list_tables to see what exists, then use describe_table on each table name. This confirms that the correct columns are present and formatted as VECTORs before writing any code.
Quick Data Auditing
A DBA needs to know if a core customer table is nearing capacity. They don't query a dashboard; they ask their agent to run table_stats on the 'CUSTOMERS' table, getting immediate row counts and optimizer data.
Debugging Index Setup
A developer suspects a vector search is failing. They use list_vector_indexes first to confirm that HNSW indexes exist where they should be, then use get_database_version to verify the database version supports the required features.
The Tradeoffs
Treating everything like a simple SQL query
The user tries to run all similarity searches using only execute_sql_query because it's the only function they know. This fails because vector distance requires specialized syntax and functions.
→
If you need vector search, use the dedicated vector_search tool call. Only use execute_sql_query for traditional data retrieval or aggregation.
Overlooking schema changes
A developer assumes a column exists based on memory and writes a query that fails because the table structure was recently changed.
→
Always run describe_table first. This confirms the current column name, data type, and whether it's VECTOR-enabled before you try to access it.
Ignoring version compatibility
A user attempts advanced vector operations without knowing if their specific Oracle instance supports them.
→
Before running any complex query, run get_database_version. This confirms the 23ai feature set is fully available for reliable development.
When It Fits, When It Doesn't
Use this server if your data access requires handling both standard relational queries (using execute_sql_query) AND specialized vector similarity searches (using vector_search). That's the core requirement. If you only need to write simple SELECT statements without any embedding or indexing logic, a basic SQL client is fine—you don't need the full complexity of this server. Conversely, if your data access never leaves plain JSON objects and involves no structured database interaction, then this toolset is overkill. You must be dealing with an actual relational vector store in Oracle 23ai.
Common Questions About Oracle Vector DB MCP
How do I check if my current Oracle version supports vector functions using get_database_version? +
Run get_database_version. This tool gives the exact runtime banner, confirming if key features like VECTOR data types and VECTOR_DISTANCE are available before you start building queries.
What is the difference between execute_sql_query and vector_search? +
execute_sql_query runs standard SQL for retrieving records based on criteria (like WHERE name='X'). vector_search uses specialized algorithms to find records that are semantically similar, even if their names don't match.
How can I see what tables are available in my schema? Use list_tables. +
Use the list_tables tool. It immediately lists every table accessible by your connected user, giving you a complete map of data sources you can query.
Should I use describe_table before running vector_search? +
Yes. Use describe_table first to confirm that the specific column you plan to search is indeed defined with the VECTOR data type, preventing runtime errors.
I need to check which specialized AI Vector search indexes are available; how do I use list_vector_indexes? +
The tool lists all instantiated vector indexes, including HNSW and IVF. It confirms where your embedding data is optimized for fast searching, preventing the need to manually track index names across multiple tables.
Before running a complex search, how do I check table cardinality and performance readiness using table_stats? +
This tool provides crucial metrics like row counts and optimizer statistics. You use it for capacity planning or tuning queries; knowing the table's current size helps predict query performance accurately.
What limitations should I be aware of when using execute_sql_query? +
You must restrict the results fetched to maintain stability. The tool warns that payload size is limited, so always include a clause like 'FETCH FIRST 100 ROWS ONLY' in your SQL statement.
When running vector_search, what distance metrics can I specify (cosine vs Euclidean)? +
You select the appropriate metric within the search parameters. Cosine and Euclidean distances measure similarity differently; choosing correctly ensures that the returned neighbors are truly relevant to your embedding.
Does it work with Oracle Autonomous Database? +
Yes. Oracle Autonomous Database on OCI has ORDS enabled by default. Use the ORDS URL from your ADB instance (e.g., https://xxxxx.adb.us-ashburn-1.oraclecloudapps.com/ords), your schema name (typically ADMIN), and the admin credentials. The VECTOR type is available on all 23ai-compatible instances.
Can I run arbitrary SQL — not just vector searches? +
Yes. The execute_sql_query tool accepts any valid Oracle SQL statement and returns results through ORDS. Add FETCH FIRST N ROWS ONLY to keep payloads manageable. This makes the agent useful for relational queries too, not just vector operations.
Which distance metrics are available for vector search? +
Oracle 23ai supports COSINE and EUCLIDEAN (L2) distance metrics natively via VECTOR_DISTANCE. Specify the metric when running a search — cosine is recommended for most text embedding use cases, while L2 works better for image or audio embeddings.
Use it with your favorite AI tools
Connect this server to Cursor, Claude, VS Code, and more.