Snowflake shares best practices and tools for its CoCo AI coding agent
Snowflake provides a comprehensive guide on leveraging its CoCo AI coding agent for data engineering. The article outlines best practices for using AI agents to achieve reproducible, high-quality results, distinguishing professional data engineers. It highlights CoCo's unique advantages for Snowflake users, including in-platform inference, native integration, and extensive built-in skills. Data engineers can access CoCo via CLI, a VS Code-based desktop IDE, or directly within Snowsight, and build consistent workflows using open-standard skills and plugins.
- →Best practices for data engineering with AI coding agents
- →Advantages of Snowflake's CoCo AI coding agent
- →Accessing Snowflake CoCo via CLI, Desktop, or Snowsight
- →Building reproducible workflows with CoCo skills and plugins
Notes (4) ›
- Best practices for data engineering with AI coding agents
Data engineers should start minimally, understand model intelligence, and prioritize conciseness and reproducibility when using AI coding agents. It's crucial to remember that agents are not enterprise data engineering tools and should never make direct changes to production environments due to their non-deterministic nature.
- Advantages of Snowflake's CoCo AI coding agent
CoCo is optimized for Snowflake data engineers, performing inference within Snowflake's security boundaries and offering deep native integration with schemas, tables, and query history. It includes numerous built-in skills tailored for Snowflake native workflows and provides flexibility in choosing top-tier frontier models.
- Accessing Snowflake CoCo via CLI, Desktop, or Snowsight
CoCo can be accessed through a command-line interface for terminal-native engineers, a VS Code-based desktop IDE for local repository integration, or directly within Snowsight without installation. All three environments share largely similar functionalities for Snowflake native intelligence.
- Building reproducible workflows with CoCo skills and plugins
Data engineers can achieve reproducible results by encoding expertise into artifacts like AGENTS.md, custom agents, slash commands, hooks, and especially Skills and Plugins. Skills implement multi-step workflows using an open standard (agentskills.io), while Plugins provide advanced packaging and deployment for skills, agents, and other components, ensuring versioning and validation within the Snowflake Plugin Catalog.
https://www.snowflake.com/content/snowflake-site/global/ja/blog/snowflake-coco-data-engineering
Related releases
- Snowflake simplifies high-throughput data streaming to Apache Iceberg with Snowpipe Streaming Snowflake Blog ·
- Snowflake Cortex AI Integrates Grok 4.6 for Advanced Agentic AI Capabilities Snowflake Blog ·
- Snowflake Cortex AI Now Supports Anthropic's Claude Fable 5.1 Snowflake Blog ·
- Snowflake on Why Business Context is Critical for Effective AI Strategies Snowflake Blog ·
- Snowflake outlines a 5-stage maturity model for autonomous data engineering Snowflake Blog ·
- Snowflake Announces GA of User-Level Quotas for AI and Compute Spend Snowflake Blog ·