Snowflake MCP. Map schemas and run queries without leaving your IDE.
Gemini Works with every AI agent you already use
…and any MCP-compatible client
Just plug in your AI agents and start using Vinkius.
Snowflake connects your AI agent directly into a live Snowflake Data Cloud. Use it to map complex schemas, list all databases and tables, audit compute costs, or run read-only SQL queries without leaving your IDE.
It lets you talk to the data warehouse like a native developer.
What your AI agents can do
Execute sql
Runs a SQL query against Snowflake, favoring read-only statements when possible.
Get query status
Checks the current status of any long-running or asynchronous SQL query.
List databases
Returns a list of all databases available within your Snowflake account.
Runs read-only SQL statements against your connected Snowflake account.
Lists every database, schema, and table available in the entire Snowflake environment.
Retrieves a list of all active virtual warehouses to check current compute costs and status.
Checks the real-time status of any lengthy, asynchronous query you've started.
Lists all internal and external stages where raw data loads or files reside.
Ask AI about this MCP
Supported MCP Clients
OAuth 2.0 Compatible Gemini Waiting for input…
Snowflake: 7 Tools for Data Architecture
Use these seven tools to query data, discover the full schema structure, and manage warehouse resources within your IDE.
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 Snowflake on Vinkius019d760aexecute sql
Runs a SQL query against Snowflake, favoring read-only statements when possible.
019d760aget query status
Checks the current status of any long-running or asynchronous SQL query.
019d760alist databases
Returns a list of all databases available within your Snowflake account.
019d760alist schemas
Lists all schemas contained within a specific database you select.
019d760alist stages
Retrieves names of internal or external data stages used for loading files.
019d760alist tables
Lists all tables that exist within a specific schema you target.
019d760alist warehouses
Outputs a list of virtual compute warehouses and their current running status.
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 Snowflake, 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 Snowflake. 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.
Juggling tabs to check basic schema details shouldn't take five minutes.
Right now, checking if a table exists or finding the right column type means jumping between your IDE, the Snowflake web interface, and maybe even consulting documentation. You copy-paste credentials here, switch tabs there, then run one query, only to find out you missed a required schema layer.
With this MCP server, your agent handles all that complexity in one chat window. Tell it what data you need—say, 'Give me the top 5 sales records.' The agent maps the databases using `list_databases`, locates the correct tables via `list_tables`, and runs the query with `execute_sql`. You just get the answer.
Snowflake MCP Server: Run any data operation directly from chat.
You no longer have to manually remember which listing tool applies where. Need to find a staging file? Use `list_stages`. Want to know if the compute cluster is running too much? Hit up `list_warehouses` first. It’s systematic.
What changes is that your AI client treats the entire data cloud—every database, every table, every stage—as one connected resource pool accessible via simple commands. That's how fast you work now.
What you can do with this MCP connector
You're connecting your AI agent straight into a live Snowflake Data Cloud. You use this server to map out complex schemas, list every table and database name, audit compute costs, or run read-only SQL queries without ever leaving your IDE. It lets you talk to the data warehouse like an actual developer.
To figure out what's running in your account, you first call list_databases to get a roster of all available databases. From there, you can narrow it down by calling list_schemas, which outputs every schema housed inside that specific database. Once you've got the right scope, you use list_tables to pull up a list of every table sitting within that exact schema.
This sequence lets your agent walk you through the entire data structure, making sure you can pinpoint any column.
To check on the infrastructure doing all this work, you run list_warehouses. It gives you a clean rundown of every virtual compute warehouse and what its current running status is; that helps you track compute usage and costs. If your raw data files are loaded using stages, you use list_stages to retrieve names for both internal and external data storage areas.
The main action is executing code. You tell the agent to run a query with execute_sql. This runs SQL statements against your connected account, favoring read-only actions whenever it can. When you fire off a complex or lengthy statement, you don't just forget about it; you use get_query_status to check the real-time status of that asynchronous job until it’s done.
This tool set means your agent doesn't guess—it validates every command against the live engine before running it. You can ask it to run a complex query based on human input or test intricate table joins safely right inside Cursor or Claude, knowing it'll only interact with read-only data until you tell it otherwise.
019d760a-a4c4-72da-b8b9-a40866890fe6 
How Snowflake MCP Works
- 1 Subscribe to the server, then provide your specific Snowflake Account identifier (e.g.,
abc123.us-east-1). - 2 Inject your necessary Snowflake OAuth token or JWT authentication key pair.
- 3 Ask your AI client (Claude or Cursor) to look into a schema using natural language prompts.
The bottom line is, you get full data visibility and query execution right inside your coding environment, ditching the browser tabs.
Who Is Snowflake MCP For?
Data Engineers who are tired of context switching between their IDE and a web dashboard. Analytics Engineers who need to validate complex joins against live definitions. Software Architects needing raw diagnostic query metrics without downloading heavy SDKs.
Validates that raw datasets correctly land in internal environments using list_stages directly from their editor.
Generates highly accurate SQL modeling by letting the agent examine live table definitions using list_tables.
Writes an agent script to pull raw diagnostic query metrics and warehouse status without needing a local SDK setup.
What Changes When You Connect
- See the full data structure by using
list_databasesfollowed bylist_schemas. You map out the entire data landscape before writing a single line of code. - Audit compute costs instantly. Use
list_warehousesto see which virtual clusters are running, helping you manage spending without manual dashboard checks. - Test complex joins safely. You can prompt your agent with an SQL query and let it run against the live engine using
execute_sql, validating results immediately. - Keep track of big jobs. If a data pipeline runs for hours, don't worry. Use
get_query_statusto monitor its progress asynchronously. - Validate raw file uploads. Need to know where external files landed? Run
list_stagesto see all internal and external staging areas.
Real-World Use Cases
Debugging Data Flow Discrepancies
A data engineer notices a table seems empty. Instead of checking the UI, they prompt their agent: 'What stages are available for this database?'. The agent calls list_stages and shows the exact path where files should be landing, solving the missing data problem instantly.
Modeling a New Data Source
An analytics engineer needs to join three tables but isn't sure which ones exist. They ask their agent to map the structure: 'List all schemas in MASTER_DB and then list the tables within ANALYTICS_GOLD.' The agent calls list_schemas and then list_tables, giving them a complete, joined reference guide for modeling.
Performance Troubleshooting
The team is running slow queries. A software architect asks the agent to check resource usage: 'What warehouses are active right now?' The agent runs list_warehouses, showing a cluster that's idle but still incurring costs, allowing the architect to shut it down.
Validating ETL Pipeline Integrity
A developer just loaded a massive dataset. They need confirmation of where the raw files landed. They use list_stages to confirm the exact internal stage path, verifying that the data source was ingested correctly before proceeding with transformations.
The Tradeoffs
Guessing Schema Depth
Writing a query like SELECT * FROM master.customers.transactions because you think it exists, only to get an 'Object not found' error.
→
Never guess. Start by asking the agent: 'What databases do I have?' Use list_databases, then drill down with list_schemas, and finally confirm the table name using list_tables.
Over-relying on UI Clicks
Switching between Snowflake Snowsight, their local SQL client, and a documentation site just to validate one column type.
→
Keep everything in the chat. Use list_tables to see the schema definitions live, keeping your cursor exactly where it is.
Forgetting Status Checks
Running a huge query and assuming it's done because the prompt returns a confirmation message.
→
Always verify long jobs. After running a major job, follow up with 'What is the status of that last query?' to use get_query_status.
When It Fits, When It Doesn't
Use this server if your core problem involves data discovery or querying massive datasets stored in Snowflake. Specifically, you need a reliable way to map out the structure—from databases (list_databases) down to individual tables (list_tables). You should use it when your workflow requires auditing compute costs (list_warehouses), checking file locations (list_stages), or running complex joins on live data using execute_sql.
Don't use this if you only need to manage user roles, run stored procedures that aren't query-based, or interact with the operational features of Snowflake (like setting up alerts). For those tasks, you'll need a dedicated administrative tool. This server is purely for reading, mapping, and executing standard SQL queries against data.
Common Questions About Snowflake MCP
How do I find all databases using list_databases? +
It's straightforward: just ask your agent to run list_databases. It returns a clean, complete list of every database in your account. This is the starting point for any data mapping effort.
If I forget a table name, can I use list_tables? +
Yes. If you know the schema but not the exact table, prompt the agent to run list_tables within that specific schema. It gives you all available tables so you don't have to guess.
What if my query is too complex for execute_sql? +
If your query involves multiple steps or runs for a long time, run it first with execute_sql and then follow up by asking the agent to check its status using get_query_status. This keeps you informed.
How do I see if my compute costs are getting too high? +
Run list_warehouses. This tool lists all virtual warehouses, letting you immediately see which clusters are running or suspended. You can use this to manage and control your spending.
What credentials do I need for my AI client to use `execute_sql`? +
You must provide a valid Snowflake OAuth token or JWT pair. The system needs explicit access rights and your account identifier (e.g., abc123.us-east-1) to run any queries against the live data.
When should I use `list_stages` instead of just querying a table? +
Use list_stages when you need to locate external or internal cloud storage pointers. This tool shows where raw, unmodeled source files actually sit before they enter the database structure.
If my query takes hours, how do I track its progress using `get_query_status`? +
After running a large job with execute_sql, use get_query_status and provide the unique query ID. This retrieves real-time status updates so you don't have to wait for a timeout.
How does `list_schemas` help me validate column names before I write a JOIN? +
list_schemas helps you map the entire structure, allowing your agent to traverse databases and identify specific schemas. This confirms that related tables exist in proximity, making joins safer.
Can my AI actually read the raw table rows via an execute statement? +
Yes. When the AI uses execute_sql with something like SELECT * FROM schema.users LIMIT 10, the MCP integration parses the exact row outputs. The LLM consumes the tabular data back into context so you can converse naturally about the dataset findings.
Is it completely safe to give AI power over a Data Warehouse? +
Safety stems from principle of least privilege. Supply a Snowflake Token tied strictly to a read-only role or a heavily scoped down service account. This allows the AI to navigate schemas and extract data without risking destructive schema mutations like DROPs or DELETEs.
Can it search for a column name if I don't know the exact schema? +
Yes! Tell your agent: 'Find which table in the SALES_DB database has a column named customer_churn_score'. Due to its autonomous workflow, the bot will pull schemas, subsequently loop over list_tables, query Snowflake’s internal information_schema if necessary, and deduce it entirely for you.
Use it with your favorite AI tools
Connect this server to Cursor, Claude, VS Code, and more.