Google has released Antigravity, an agent-integrated development environment built on VS Code. For data teams using dbt, its architecture presents specific engineering considerations. The system is a fork of VS Code, ensuring extension compatibility while layering on an agent-first design. This moves beyond code completion to planning and execution agents that operate with project-wide context.

Integration with dbt is handled through the official extension, providing lineage and query previews. The more significant design choice is the support for Model Context Protocol (MCP) servers. Connecting the dbt MCP server grants the local agent deeper awareness of project structures, enabling it to generate models, tests, and documentation. This shifts the tool from an editor to a participant in defined workflows, potentially linking with systems like BigQuery, Dataplex, and GitHub for tasks from SQL generation to pull requests.

Practically, engineers must manage this capability. Defining agent rules in configuration files for naming or SQL style acts as guardrails. Breaking large tasks into sequenced, smaller jobs improves agent output reliability. The environment also suggests a split in tool use: Antigravity for multi-agent planning and architectural work, paired with the Gemini CLI for focused, terminal-based execution. The trade-off is embedding more project logic into agent configuration versus writing code directly. For teams standardized on dbt and Google's cloud ecosystem, Antigravity reduces friction across development phases, but it requires structured prompts and rules to produce consistent, production-ready artifacts.

Source: dbt Labs Blog