Snowflake MCP server and Dome
Let agents search your Snowflake data, and decide which columns they get. Connect Snowflake's managed MCP server to Dome with Snowflake OAuth. Each call runs with the default role of the person the agent acts for. Rules read every call's arguments, down to the columns a search returns.
Dome provides a tool gatewayand authorization for Snowflake.
Decided on the columns
Every Cortex Search call must name the columns it wants back. Ask for customer_email and the call is refused.
Their role, their data
With per-user OAuth, each call runs with the default role of the person the agent acts for, and Snowflake grants still apply.
No raw SQL
The SQL execution tool is refused for agents that don't need it. They answer questions through Cortex Search and Cortex Analyst.
Snowflake behind the Gatewayin three steps.
Add your MCP server's URL, sync its tools into the Gateway's catalog, and apply the rules. The URL and the tool names come from your CREATE MCP SERVER statement.
Add the Snowflake MCP server
Create a Snowflake OAuth security integration with https://api.domesystems.ai/oauth/callback as its redirect URI. Dome holds each user's tokens and injects them on calls made for them.
shelldome tools add --name snowflake \ --url https://myorg-myaccount.snowflakecomputing.com/api/v2/databases/SUPPORT_DB/schemas/MCP/mcp-servers/support_mcp \ --auth-method oauth --credential-type per-user \ --oauth-authorize-url https://myorg-myaccount.snowflakecomputing.com/oauth/authorize \ --oauth-token-url https://myorg-myaccount.snowflakecomputing.com/oauth/token-request \ --oauth-client-id "$SNOWFLAKE_CLIENT_ID" \ --oauth-client-secret "$SNOWFLAKE_CLIENT_SECRET" \ --gateway data-gatewaySync the catalog
Run it once your own Snowflake consent is attached. Until then the sync fails, and agents get “tool not available in this gateway”.
shelldome tools catalog sync snowflakeApply the rules
Scope them to one agent while you try them, and simulate before you deploy.
shelldome rules apply snowflake-columns.cedar \ --agent churn-researcher --name snowflake-columns
Search,without the email column.
The permit opens Snowflake to the agent. The forbids refuse a search that doesn't name its columns or asks for customer_email, a search for more than 50 rows, and any call to the SQL execution tool.
cedar
permit (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "snowflake" };
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "snowflake" && resource.tool_name == "ticket-search" && resource has arguments && (!(resource.arguments has columns) || resource.arguments.columns.contains("customer_email"))};
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "snowflake" && resource.tool_name == "ticket-search" && resource has arguments && resource.arguments has limit && resource.arguments.limit > 50};
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "snowflake" && resource.tool_name == "sql_exec_tool"};One call,two outcomes.
Switch the caller or the argument and watch the same call decide differently, with every decision landing in audit.
Columns
agent churn-researcher · acting as r.chen
snowflake/ticket-search(query: "cancel after price change", columns: ["ticket_id", "subject", "status"], limit: 20)
- Agentchurn-researcher is registered and active
- Callerr.chen verified through Okta
- RuleTicket fields are allowed
DecisionAllowed
Three agentson Snowflake.
These come from our template library, and each one reads broadly and writes narrowly.
support-assistant
Find similar tickets
Searches past support tickets for the one in front of an agent and suggests the fix that worked.
finance-analyst
Answer revenue questions
Asks Cortex Analyst revenue questions in plain language and gets back the SQL that answers them.
churn-researcher
Research churn
Reads cancellation tickets alongside revenue by segment. Customer contact details stay out of its results.
Bringing it alltogether.
Connecting Snowflake to registered agents, models, and identity in Dome completes a governed agent application.
Acting for
Agent
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Tools
SnowflakeMCP server
Models
Common questionsabout Snowflake.
More of whatDome works with.
Model
Claude Fable
Fable 5.1 from Anthropic and Amazon Bedrock in one failover pool, open to one group and capped by quota.
Provider
Anthropic
The Claude API behind the Model Broker: the key held in Dome, every call authorized, metered and audited.
Identity
Okta
Okta tokens verified on every agent call, so rules and audit name the person each agent acted for.
Runtime
OpenAI Agents SDK
OpenAI Agents SDK agents with their models on the Model Broker and their MCP tools on the Tool Gateway.
Client
Claude Code
Claude Code on a Dome Gateway with per-developer sign-in, rules on every tool call, and audit by name.
Agent service
TinyFish
TinyFish's web agents behind the Tool Gateway, with rules that decide each run on the site it targets.
Talk with ourFDE team.
Or start free, and register your first agent in minutes. Connect it through a gateway, give it rules, and see every call it makes in one audit trail.
Talk with our FDE team
Our engineers design, build and run governed agents for the part of your business you want to change. You see what they deliver.
MCP servers for AI agents: per-argument rules on every call
Put any remote MCP server behind the Tool Gateway. Agents reach it through one governed endpoint, and rules decide each call on its arguments.