Skip to content

Working with clients? Stay ahead of what they ask for.

Snowflake MCP

by Snowflake-Labs

The official Snowflake MCP server. Cortex AI access, object management, SQL orchestration, and semantic view consumption. 279 GitHub stars and 2 commits on main in the last 30 days.

★ 279·7 listed tools·Released AUG 2025·Apache-2.0
pip install snowflake-labs-mcp
Share:

INSTALL THIS SERVER

Requires authenticationSnowflake account credentials (account, user, password or PAT, role, warehouse). Service-account credentials recommended for agent workflows.
{ "mcpServers": { "snowflake": { "command": "python", "args": [ "-m", "snowflake_labs_mcp" ], "env": { "SNOWFLAKE_ACCOUNT": "your-account.region", "SNOWFLAKE_USER": "agent-service-user", "SNOWFLAKE_PASSWORD": "YOUR_PASSWORD", "SNOWFLAKE_ROLE": "AGENT_READ_ROLE", "SNOWFLAKE_WAREHOUSE": "AGENT_WH" } } } }
PrereqPyPI: `snowflake-labs-mcp`. Use a dedicated service-account user with a role scoped to the agent's databases. Path: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS).
{ "mcpServers": { "snowflake": { "command": "python", "args": [ "-m", "snowflake_labs_mcp" ], "env": { "SNOWFLAKE_ACCOUNT": "your-account.region", "SNOWFLAKE_USER": "agent-service-user", "SNOWFLAKE_PASSWORD": "YOUR_PASSWORD", "SNOWFLAKE_ROLE": "AGENT_READ_ROLE", "SNOWFLAKE_WAREHOUSE": "AGENT_WH" } } } }
{ "mcpServers": { "snowflake": { "command": "python", "args": [ "-m", "snowflake_labs_mcp" ], "env": { "SNOWFLAKE_ACCOUNT": "your-account.region", "SNOWFLAKE_USER": "agent-service-user", "SNOWFLAKE_PASSWORD": "YOUR_PASSWORD", "SNOWFLAKE_ROLE": "AGENT_READ_ROLE", "SNOWFLAKE_WAREHOUSE": "AGENT_WH" } } } }
{ "mcpServers": { "snowflake": { "command": "python", "args": [ "-m", "snowflake_labs_mcp" ], "env": { "SNOWFLAKE_ACCOUNT": "your-account.region", "SNOWFLAKE_USER": "agent-service-user", "SNOWFLAKE_PASSWORD": "YOUR_PASSWORD", "SNOWFLAKE_ROLE": "AGENT_READ_ROLE", "SNOWFLAKE_WAREHOUSE": "AGENT_WH" } } } }
{ "mcpServers": { "snowflake": { "command": "python", "args": [ "-m", "snowflake_labs_mcp" ], "env": { "SNOWFLAKE_ACCOUNT": "your-account.region", "SNOWFLAKE_USER": "agent-service-user", "SNOWFLAKE_PASSWORD": "YOUR_PASSWORD", "SNOWFLAKE_ROLE": "AGENT_READ_ROLE", "SNOWFLAKE_WAREHOUSE": "AGENT_WH" } } } }

7 SELECTED TOOLS

execute_query
Run a parameterised SQL statement against the warehouse
Admin
list_databases
List databases visible to the connected role
Read
list_schemas
List schemas within a database
Read
describe_table
Inspect columns and types for a table or view
Read
cortex_search
Run a Cortex Search service query for retrieval-augmented workflows
Read
cortex_complete
Run a Cortex Complete LLM completion
Read

OUR ASSESSMENT

Snowflake-Labs/mcp ships the official Snowflake MCP server. Tool surface covers SQL execution, database and schema discovery, Cortex Search and Cortex Complete (Snowflake LLM and retrieval), and semantic view listing. Distinguishes from a generic Postgres MCP by exposing Cortex AI primitives and Snowflake-specific governance metadata.

Strengths
  • 279 GitHub stars and Apache-2.0 license.
  • Snowflake official maintenance under the Snowflake-Labs org.
  • Cortex Search and Cortex Complete tools expose Snowflake LLM and retrieval inside the agent.
  • Semantic view listing surfaces governed analyst views, not just raw tables.
  • 2 commits on main in the last 30 days.
Weaknesses
  • Snowflake account required; not a self-hosted MCP.
  • Cortex AI tools incur per-query Snowflake compute cost beyond the underlying warehouse.
  • Service-account permissioning is the operator responsibility; default role can over-grant.
Security Notes

Use a dedicated service-account user with a role scoped to the specific databases the agent should reach. Cortex Complete and execute_query carry compute cost; set per-warehouse spending limits in the Snowflake admin console.

Best For

Snowflake-first analytics teams adopting agentic data workflows; Cortex AI users who want LLM completions inside the warehouse exposed to the agent; teams using semantic views for analyst-friendly query surfaces.

TECHNICAL DETAILS

Language
python
Transport
stdio
Clients
Claude DesktopClaude CodeCursorVS CodeWindsurf
License
Apache-2.0
GitHub
npm
snowflake-labs-mcp
Last Release
snowflake-labs-mcp (PyPI latest)MAY 5, 2026
First Released
AUG 1, 2025

RECORDED ADOPTION METRICS

Saved directory snapshots. Record last checked MAY 5, 2026. The review date above does not refresh these figures.

// GitHub Stars
279

// Recorded note279 stars on Snowflake-Labs/mcp. 2 commits on main in the last 30 days.

// Popularity Rank
#2
Recorded directory rank · #2 in Data / Analytics

// Recorded notePairs with Tableau, Power BI, dbt, Google Analytics MCPs in the data category. Cortex AI tools differentiate from generic SQL MCPs.

SOURCES & RECORDED CHECKS

These are the directory listings and check details held in this record. Client compatibility, directory listings and installation testing are separate signals. Recorded client names describe the scope of a check; they do not establish a security assessment or a recurring test schedule.

01
Discovered
Source not recorded
First indexed MAY 5, 2026
02
Directory listings
5 directories
PulseMCP, MCP.so, Glama, Smithery, Official MCP Registry
03
Recorded client testing
Claude Desktop, Cursor
Check current setup against the provider documentation
04
Record last checked
MAY 5, 2026
Record date; functional retest scope may differ
// How other directories see it

The same server, 5 different lenses. We reconcile these signals into our editorial score, which is why our number sometimes diverges from a directory-aggregate star count.

SourceTheir ratingTheir star countTheir downloadsLast synced
AutomationSwitch This page4.4editorial279—MAY 5, 2026
PulseMCP— unratedunavailableunavailableMAY 5, 2026
MCP.so— unratedunavailableunavailableMAY 5, 2026
Glama— unratedunavailableunavailableMAY 5, 2026
Smithery— unratedunavailableunavailableMAY 5, 2026
Official MCP Registry— unratedunavailableunavailableMAY 5, 2026

// Counts are directory-reported; we don't adjust them. Discrepancies usually come from different snapshot times or star-caching.

// Get in touch

DISCUSS YOUR
MCP REQUIREMENTS.

Evaluating a server, scoping an internal deployment, or working out whether MCP is the right fit at all. Start the conversation and we will point you at the right piece of the ecosystem.

Discuss Your MCP Requirements →