A programmable decision layer for Snowflake
Policies, meaning, and access rules as versioned operational metadata. Travels with your data, never leaves your Snowflake account.
The decision layer Cortex Agents already speak.
Cortex Code, Cortex Agents, and Cortex Analyst call the same Metatate tools — through MCP or as native Snowflake SQL functions. Same engine, two surfaces, every call inside your Snowflake account.
Metatate inside the Snowflake-native developer workspace
Cortex Code is Snowflake's coding agent. The Metatate plugin adds slash commands and a grounding skill, talking to the Snowflake-managed MCP server installed by the Native App.
- Snowflake-managed MCP — zero data egress, audit stays in account
- Role-restricted PAT authentication — no OAuth browser flow
- Slash commands across discover, decide, validate, and release-gate
How It Works
Author decision-grade instructions, deploy context layers, and expose structured metadata to agents and workflows.
Define Policy Identity
Define policies, business logic, access controls, and data meaning as structured metadata. The AI assistant compiles plain-language descriptions into machine-readable instructions with conditions, scoped constraints, and compliance tags.
Apply Meaning Layer
Apply context layers to tables, fields, or entire workflows. Permitted and prohibited uses, ownership, source provenance, and AI/ML governance — persistent context that travels with your data, not in external documentation.
Decision Audit Trail
Agents and applications query context programmatically. Every policy change and tool call is captured — agents validate intent and read constraints before they act.
Operational metadata as a decision layer
Structured metadata that agents can parse, query, and act on. Context moves with data, not in external documentation.
Persistent Context
Business logic, data contracts, policies, source provenance, and access rules as structured, machine-readable context. Versioned metadata that evolves with your data.
Programmable Rules
Agents evaluate dynamic business rules instead of relying on hardcoded branches or human interpretation. Define conditional rules, usage constraints, and compliance tags.
Agent-Ready MCP Server
Agents query and reason over structured context through the Model Context Protocol. Works with Snowflake Intelligence — agents understand meaning, not just structure.
Governance Without Bottlenecks
Agents check constraints before they act, so domain expertise reaches every decision without a review queue.
Same engine. Two access surfaces. One Snowflake account.
The Metatate Native App exposes its decision tools through both an MCP HTTP server and as native Snowflake SQL functions. Cortex calls either — everything resolves inside your account.
Instructions, policies, and column-level meaning as versioned operational metadata. Plan/apply workflow. Diffable. Reviewable.
Tables, views, and pipelines stay in your Snowflake account. Metatate references them through standard Snowflake objects — no replication, no shadow store.
Every layer above lives inside your Snowflake account. No external API gateway. No tokens crossing your perimeter. Cortex Agents call Metatate the same way they call any other Snowflake-native function.
Zero Data Egress
All processing stays inside your Snowflake account. Context travels with data — no external API calls, no data leaving your perimeter.
Snowflake Cortex AI
Compiles decision-grade instructions using Snowflake's built-in AI. No external API keys, no third-party inference.
Versioned & Portable
Context layers are versioned, allowing policies to evolve over time. Deploy once, version and evolve — no rigid implementations that break with updates.
Persistent context.
Consistent decisions.
Install Metatate and go from zero to governed in minutes. Same engine for analytics, AI, and operational workflows.
Get it on Snowflake Marketplace