Snowflake
Cloud data warehouse.
Verdict
Common use cases
- Run ad-hoc SQL queries from chat
- Cancel long-running statements during incidents
- Check maintenance windows before deployments
- Monitor component health during outages
- Fetch catalog integration details for audits
Integration
- Vendor
- Snowflake
- Category
- developer-tools
- Auth
- OAUTH2
- Tools
- 15
- Composio slug
snowflake
Tools
- Cancel Statement Execution
Cancels the execution of a running sql statement. use this action to stop a long-running query.
- Check Statement Status
Retrieves the status of a previously submitted sql statement. replace '{statementhandle}' with the unique identifier of the statement.
- Execute SQL
Tool to execute a sql statement and return the resulting data. use when you need to query data from snowflake.
- Fetch Catalog Integration
Fetches details of a specific catalog integration.
- Get Active Scheduled Maintenances
Retrieves a list of any active scheduled maintenances currently in the in progress or verifying state.
- Get All Scheduled Maintenances
Retrieves a list of the 50 most recent scheduled maintenances, including those in the completed state.
- Get Component Status
Retrieves the status of individual components, each listed with its current status.
- Get Status Rollup
Retrieves the status rollup for the entire page, including indicators and human-readable descriptions of the blended component status.
- Get Status Summary
Retrieves a summary of the status page, including status indicators, component statuses, unresolved incidents, and upcoming or in-progress scheduled maintenances.
- Get Unresolved Incidents
Retrieves a list of any unresolved incidents currently in the investigating, identified, or monitoring state.
- Get Upcoming Scheduled Maintenances
Retrieves a list of any upcoming scheduled maintenances still in the scheduled state.
- Show Databases
Lists all databases for which you have access privileges. shows database metadata including name, creation date, owner, retention time, and more. can filter results and include dropped databases within time travel retention period.
- Show Schemas
Lists all schemas for which you have access privileges. shows schema metadata including name, creation date, owner, database, retention time, and more. can filter results and include dropped schemas within time travel retention period.
- Show Tables
Lists all tables for which you have access privileges. shows table metadata including name, creation date, owner, database, schema, row count, size in bytes, clustering keys, and more. can filter results and include dropped tables within ti
- Submit SQL Statement
Submits a sql statement for execution.
Setup
Setup guide
- 11. In your Switchy workspace, click Integrations in the left sidebar, then find Snowflake under Developer Tools. 2. Click Connect and you'll redirect to Snowflake's OAuth consent screen. 3. Sign in with your Snowflake account credentials (the account that has access to the warehouses you want to query). 4. Review the requested scopes: read access to databases and schemas, execute permissions for SQL statements, and status monitoring for maintenance windows. 5. Click Authorize to grant access, then Switchy redirects you back with a confirmation message. 6. Open any Space and type '@Snowflake execute SQL: SELECT CURRENT_TIMESTAMP();' to verify the connection works. 7. If the query returns a timestamp, you're ready — the MCP can now run queries, check statement status, and fetch maintenance schedules on your behalf.
What teammates see: by default, memories from Snowflake are scoped to the Space (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Spaces have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Query Recent Orders
@Snowflake execute SQL: SELECT order_id, customer_id, total_amount, created_at FROM sales.orders ORDER BY created_at DESC LIMIT 100;Open in a Space →
Cancel Runaway Query
@Snowflake cancel the statement with handle '01a2b3c4-5678-90de-f123-456789abcdef' — it's been running for 20 minutes and blocking our ETL job.Open in a Space →
Check Query Status
@Snowflake check the status of statement handle '01a2b3c4-5678-90de-f123-456789abcdef' and tell me if it's still running or if it finished.Open in a Space →
Review Active Maintenance
@Snowflake get all active scheduled maintenances right now — we're about to deploy and need to know if anything's in progress.Open in a Space →
Platform Health Check
@Snowflake get the status rollup and component statuses — our queries are timing out and I want to rule out platform problems.Open in a Space →
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@snowflake run a query to show me the top 10 customers by total revenue in the last quarter, grouped by region
Executed SQL query against your Snowflake warehouse. Results show EMEA leading with $2.4M across 3 customers (Acme Corp, GlobalTech, EuroSystems), followed by North America at $1.8M from 4 customers, and APAC at $1.1M from 3 customers. Query completed in 1.2 seconds using the COMPUTE_WH warehouse. Full result set contains 10 rows with columns: REGION, CUSTOMER_NAME, TOTAL_REVENUE, ORDER_COUNT.
This example uses the Execute SQL tool to run analytics queries. Requires OAuth scope for query execution on the target warehouse. Query performance depends on warehouse size and data volume. The MCP returns structured results that the AI can summarize — actual column names and values will match your schema.
@snowflake I started a long ETL query 20 minutes ago but need to stop it now — can you cancel statement handle 01b2a3c4-d5e6-7f8g-9h0i-1j2k3l4m5n6o?
Cancelled the running SQL statement with handle 01b2a3c4-d5e6-7f8g-9h0i-1j2k3l4m5n6o. The query was in execution state for 18 minutes before cancellation. Snowflake will roll back any uncommitted changes from this statement. You can verify cancellation status by checking the statement handle again if needed. Note that warehouse compute credits were consumed up to the cancellation point.
Demonstrates the Cancel Statement Execution tool for stopping runaway queries. Useful when a teammate kicks off an expensive operation by mistake. Requires the statement handle from the original query submission. Cancellation is not instant — Snowflake may take seconds to terminate the query depending on its current phase.
@snowflake check if there's any scheduled maintenance coming up that might affect our data pipelines this week
Retrieved active and upcoming scheduled maintenances from Snowflake status API. One maintenance window is scheduled for Thursday 2:00-4:00 AM UTC affecting the US-WEST-2 region: 'Routine infrastructure upgrades for improved query performance.' Status is currently 'scheduled' (not yet in progress). No other maintenances found in the next 7 days. This maintenance may cause brief connection interruptions or slower query execution during the window.
Uses the Get All Scheduled Maintenances tool to surface operational awareness. Helps teams plan around downtime or performance degradation. The MCP queries Snowflake's public status page, so results reflect vendor-announced maintenance only — not account-specific issues. Pair this with monitoring tools for complete visibility.
Use-case deep-dives
When Snowflake MCP helps ops teams debug outages faster
A 6-person data team gets paged at 2am because the nightly ETL job is stuck. With Snowflake MCP, the on-call engineer asks Switchy to check statement status for the last three hours, spots a runaway query consuming warehouse credits, and cancels it without logging into the console. The MCP shines here because OAuth2 means no shared service account credentials floating around Slack threads. The trade-off: if your team runs more than 20 queries per minute, the statement-handle lookup gets noisy and you're better off with direct Snowflake alerting. For small teams triaging occasional stuck jobs, this MCP turns a 15-minute console hunt into a 90-second chat exchange.
Using Snowflake MCP for ad-hoc customer data lookups
A 3-person support team fields questions about order history and account balances. They store everything in Snowflake but hate context-switching to run SQL. With this MCP, a support rep asks Switchy to execute a query for a customer's last five transactions, gets the result inline, and pastes it into the ticket. The win is speed: no VPN, no BI tool login, no waiting for the data analyst. The boundary: if your queries need complex joins or return more than 500 rows, the MCP's execute tool times out and you're back to the console. For simple lookups under 10 seconds, this MCP keeps support moving without breaking their flow.
When to use Snowflake MCP for maintenance awareness
A 4-person analytics team runs daily reports and needs to know if Snowflake is degraded before users complain. With this MCP, they set up a morning standup ritual where Switchy checks active scheduled maintenances and component status, then summarizes any yellow or red indicators. The MCP's status tools pull from Snowflake's public status API, so the team sees platform issues without manually checking status.snowflake.com. The limit: this doesn't monitor your warehouse performance or query costs, just Snowflake's infrastructure health. If you need cost alerts or slow-query detection, you still need Snowflake's native monitoring. For teams that want a quick platform health check without leaving their workspace, this MCP adds useful visibility.
Frequently asked
What can the Snowflake MCP do in Switchy?
It lets your team run SQL queries against your Snowflake data warehouse directly from Switchy's AI workspace. You can execute statements, check query status, cancel long-running queries, and monitor Snowflake's maintenance windows and component health. This means analysts can pull data without switching to Snowflake's console or writing custom scripts.
Do I need Snowflake admin access to connect this MCP?
You need a Snowflake account with OAuth configured and sufficient privileges to execute SQL in your target databases. Snowflake's OAuth2 flow will prompt you to authorize Switchy's access. If your org restricts OAuth apps, you'll need IT approval before connecting. The MCP doesn't require account-level admin rights, just query permissions on the warehouses and schemas your team uses.
Can the Snowflake MCP write data or only read it?
It can execute any SQL statement your Snowflake account allows, including INSERT, UPDATE, DELETE, and DDL commands like CREATE TABLE. The MCP doesn't artificially limit write operations. If you want read-only access, configure a Snowflake role with SELECT-only privileges and connect using that role. Otherwise, the AI can modify data if your use case requires it.
How does this compare to querying Snowflake through its web UI?
The MCP brings Snowflake queries into your team's AI workflow, so you don't context-switch to Snowflake's console. You lose the visual query editor and result-set formatting Snowflake's UI provides, but you gain the ability to chain SQL results into other MCP actions or ask the AI to interpret data inline. Best for ad-hoc analysis; use Snowflake's UI for building production dashboards.
Who on the team should connect the Snowflake MCP?
Whoever owns your Snowflake credentials and understands which databases the team needs access to. Typically a data analyst or engineer. Once connected, any Switchy workspace member can trigger queries through the AI, but the MCP runs with the permissions of the account that authenticated. If multiple people need different Snowflake roles, each should connect their own account.