The modern software development landscape isn’t happening just on one surface — it’s happening across an entire ecosystem of agentic tools. Agents are being developed at an unprecedented scale, and these agents require direct access to enterprise data for context and grounding.
However, the current tooling for building agents and managing data is heavily fragmented. This can make it difficult to access data, increasing security risks, and causing broken developer experiences that hinder innovation.
To address this challenge, we recently launched Data Agent Kit, a unified, open-source collection of data engineering and data science skills, tools and plugins that integrate directly into the environments practitioners already use, such as VS Code, Claude Code, Codex, Gemini CLI and the Antigravity CLI. By seamlessly bringing together these core tools and skills with your enterprise data, the Data Agent Kit effectively serves as a comprehensive harness for agentic context, memory, and personalization. It provides:
-
Agentic skills: Pre-codified pathways for interacting with your data estate, covering query optimization, ML best practices, data validation, data drift checks, governance, and troubleshooting.
-
Model Context Protocol (MCP) tools: Secure connections between agentic workflows and cloud data platforms like BigQuery, AlloyDB, and Google Cloud Storage. Developers can now configure connection parameters for their cloud datasets and data processing engines without having to manage complex, manual pipeline code.
-
Plugins and extensions: Native IDE integrations that enable rich, context-aware developer interactions.
Together, these Data Agent Kit capabilities help data practitioners go from manually writing code to intent-driven data science and engineering: defining the desired business outcomes, constraints, and success criteria, and allowing the AI-augmented system to figure out how to execute it. This shift is critical because today, when building agentic applications that navigate complex data architectures, there’s often a ‘context window tax’ i.e., developers have to manually paste vast amounts of schema metadata into prompts, eating up token limits and increasing latency. Meanwhile, data practitioners often lack guidance about how to efficiently query, optimize, and troubleshoot cloud data, while specialized, fragmented development environments cannot see across your entire data estate. Data Agent Kit helps with these challenges and others, providing the foundational capabilities data practitioners need for a new agentic way of working.
Read on for an overview of Data Agent Kit’s features and benefits, how to install it and connect your local environment to your data estate, and an intent-driven engineering example.
A unified hub for your data estate and lifecycle
Data Agent Kit makes your entire data estate available in a single view. This goes beyond providing a simple catalog for databases such as BigQuery, AlloyDB and Spanner; rather, it integrates data engineering and science tasks, orchestration pipelines, and jobs into a single interface. This allows practitioners to manage their entire data workflow — from discovery to production — without context switching. Data Agent Kit’s intelligent routing automatically chooses the optimal compute engine for your task — whether that’s BigQuery for SQL-native analytics and ELT, or Spark for custom Python transformations and distributed ML training.
