Skip to main content

MCP Tool Schemas

JSON input schemas for all seven MCP tools registered with the Metatate MCP server. SQL/backend APIs use underscore names, while the Snowflake MCP server exposes hyphenated names.

discover_context​

Discover governed tables. All parameters are optional.

{
  "type": "object",
  "properties": {
    "database": { "type": "string", "description": "Filter by database name" },
    "schema": { "type": "string", "description": "Filter by schema name" },
    "min_sensitivity": {
      "type": "string",
      "enum": ["PUBLIC", "INTERNAL", "CONFIDENTIAL", "RESTRICTED"],
      "description": "Minimum sensitivity level to include"
    },
    "compliance_any": {
      "type": "array",
      "items": { "type": "string" },
      "description": "Include tables matching any customer-defined compliance or control tag"
    },
    "has_pii": { "type": "boolean", "description": "Filter to tables that contain PII columns" },
    "domain": { "type": "string", "description": "Filter by business domain" }
  }
}

get_decision_context​

Retrieve governance context for a table, including policy summary, business context, lineage, retention, and rule summaries.

{
  "type": "object",
  "properties": {
    "table_name": { "type": "string", "description": "Fully qualified table name (DATABASE.SCHEMA.TABLE)" }
  },
  "required": ["table_name"]
}

inspect_data_meaning​

Inspect column-level classifications, sensitivity, masking, and PII indicators.

{
  "type": "object",
  "properties": {
    "table_name": { "type": "string", "description": "Fully qualified table name (DATABASE.SCHEMA.TABLE)" },
    "columns": {
      "type": "array",
      "items": { "type": "string" },
      "description": "Specific column names to inspect. Omit to inspect all columns."
    }
  },
  "required": ["table_name"]
}

inspect_governance_rules​

Inspect deployed usage, validation, and transfer rules with provenance for a governed table.

{
  "type": "object",
  "properties": {
    "table_name": { "type": "string", "description": "Fully qualified table name (DATABASE.SCHEMA.TABLE)" },
    "columns": {
      "type": "array",
      "items": { "type": "string" },
      "description": "Optional columns used to scope column-governance facts"
    }
  },
  "required": ["table_name"]
}

authorize_use​

Check whether a proposed operation is authorized. Natural-language requests should be normalized through /v1/mcp/authorize/assist before direct SQL execution.

{
  "type": "object",
  "properties": {
    "table_name": { "type": "string", "description": "Fully qualified table name (DATABASE.SCHEMA.TABLE)" },
    "operation": { "type": "string", "description": "Canonical operation such as read, write, train, infer, export, share, or copy" },
    "intended_use": { "type": "string", "description": "Canonical intended-use label such as analytics, reporting, marketing, ml_training, inference, external_sharing, or data_export" },
    "actor_role": { "type": "string", "description": "Role performing the operation" },
    "columns": { "type": "array", "items": { "type": "string" }, "description": "Specific columns involved in the operation" },
    "destination": {
      "type": "object",
      "description": "Destination metadata for export, share, copy, or other transfer workflows",
      "properties": {
        "system": { "type": "string", "description": "Target system name" },
        "jurisdiction": { "type": "string", "description": "Destination jurisdiction" }
      }
    },
    "consumer_jurisdiction": { "type": "string", "description": "Consumer jurisdiction for transfer governance" },
    "context": { "type": "object", "description": "Additional audit metadata logged with the decision" },
    "raw_request_text": { "type": "string", "description": "Original natural-language request for audit" },
    "normalized_request": { "type": "object", "description": "Structured normalized request for audit" },
    "normalization_meta": { "type": "object", "description": "Normalization provider, model, and confidence metadata" }
  },
  "required": ["table_name", "operation", "intended_use"]
}

Responses include decision, decision_id, rationale fields, conditions, matched instructions, and agent_action.

validate_query_context​

Validate SQL before execution. When operation is omitted, Metatate infers the operation where possible and reports extraction diagnostics.

{
  "type": "object",
  "properties": {
    "sql": { "type": "string", "description": "The SQL query to validate" },
    "operation": { "type": "string", "description": "Optional canonical operation for embedded authorization" },
    "intended_use": { "type": "string", "description": "Canonical intended-use label for embedded authorization" },
    "actor_role": { "type": "string", "description": "Role that will execute the query" },
    "destination": { "type": "object", "description": "Destination metadata for embedded transfer authorization" },
    "consumer_jurisdiction": { "type": "string", "description": "Consumer jurisdiction for embedded transfer authorization" }
  },
  "required": ["sql"]
}

Responses include validation_id, SQL findings, extracted columns, parser diagnostics, embedded authorization decisions, and agent_action.

explain_why​

Retrieve a trace for a logged authorization decision or query validation.

{
  "type": "object",
  "properties": {
    "decision_id": { "type": "string", "description": "The ID returned by authorize_use" },
    "validation_id": { "type": "string", "description": "The ID returned by validate_query_context" }
  }
}

Provide either decision_id or validation_id.