CocoPlus: A Multi-Agent Development Lifecycle Plugin for Snowflake Data Engineering

CocoPlus is a plugin for the Snowflake Cortex Code CLI that introduces structured multi-agent workflows, covering the full data engineering lifecycle from requirement definition to production deployment. It supports 8 professional roles and strict production safety controls, aiming to solve the fragmentation problem of data engineering toolchains and provide systematic support for AI-assisted development.

The data engineering field has long faced the problem of toolchain fragmentation: from SQL writing to data pipeline orchestration, test validation to production deployment, engineers need to switch tools frequently. The Snowflake Cortex Code CLI (coco) provides a unified command-line interface, but the deep integration of AI agent capabilities into workflows remains an open issue. As a plugin, CocoPlus is designed to address this gap by introducing multi-agent workflows into the full lifecycle.

The core of CocoPlus is the CocoBrew six-phase framework:

1. Spec (Requirement Definition): Convert vague business requirements into technical specifications, clarifying data models, transformation logic, and quality standards;
2. Plan (Planning): Generate implementation plans, including task decomposition, dependency relationships, and resource estimation;
3. Build (Implementation): AI-assisted SQL generation, data pipeline configuration, and transformation logic development;
4. Test (Validation): Generate and execute automated testing frameworks, covering unit tests, integration tests, and data quality validation;
5. Review (Code Review): Multi-agents check SQL quality, performance optimization, and best practices from professional perspectives;
6. Ship (Deployment): Generate deployment scripts, rollback strategies, and monitoring configurations, with the Safety Gate security layer playing a key role.

• Data Engineer ($de): Core data transformation pipelines and ETL logic;
• Analytics Engineer ($ae): Business semantics of data models and indicator definition;
• Data Scientist ($ds): Feature engineering and machine learning preparation;
• Data Analyst ($da): Data availability and query performance from the end-user perspective;
• BI Analyst ($bi): Dashboard design and KPI definition for BI scenarios;
• Data Product Manager ($dpm): Coordinate business requirements and technical implementation;
• Data Steward ($dst): Data governance, metadata management, and compliance;
• Chief Data Officer ($cdo): Strategic-level data architecture decisions. Agents work in parallel in isolated git worktrees via CocoHarvest, with unified output orchestrated through the CocoFlow JSON pipeline.

CocoPlus's Safety Gate provides three security modes:

• Strict Mode: All SQL operations require manual approval; AI-generated modification commands are not executed automatically;
• Normal Mode: AI can perform operations in a sandbox but intercepts production schema changes;
• Off Mode: Disable interception, suitable for development environments or automated CI/CD processes. The layered strategy can be flexibly configured based on environment sensitivity and team maturity.

• CocoMeter: Provides session-level and phase-level token consumption and cost tracking to help optimize agent interaction strategies;
• CocoGrove: A pattern learning library that automatically identifies and accumulates team coding patterns, naming conventions, and architectural preferences. As the project progresses, it improves the quality of suggestions, evolving from a general assistant to an exclusive advisor.

CocoPlus is open-sourced under the MIT license and is a community-driven experimental tool (not an official Snowflake product). Installation is done via npx skills add to integrate into the Cortex Code CLI. After initialization, activate the project with /pod init, and drive the six-phase workflow using commands like /spec//plan. Complete documentation is available at cocoplus.dev.

CocoPlus provides a systematic integration solution for AI agent capabilities for Snowflake data engineering teams. Its concepts of multi-role collaboration and full lifecycle management are worthy of industry attention. For enterprises using Snowflake, this tool can optimize development processes, balance innovation speed and risk control, and improve data engineering efficiency.