Google Cloud Data Agent Kit is generally available, enabling coding agents to access Google Cloud data services without manual context setup. It provides MCP tools for live environment access and open-source skills that teach agents data best practices. You can install it in VS Code, Antigravity, Claude Code, or Cloud Shell in under a minute.
Install Data Agent Kit in your IDE or terminal
To install the IDE extension, open VS Code, Antigravity IDE, Cursor, or any VS Code-compatible editor and search for 'Google Cloud Data Agent Kit' in the Extensions panel. You can also install it from the VS Code Marketplace or Open VSX. For Antigravity 2.0, go to Settings > Customizations > Build with Google Plugins and download the plugin. Cloud Shell and Cloud Workstations come with it pre-installed—just open the editor and sign in.
For terminal-based agents like Antigravity CLI, Claude Code, or Codex, use the official GitHub repository. Run 'agy plugin install https://github.com/GoogleCloudPlatform/data-agent-kit-plugin' for Antigravity CLI, 'claude plugin install data-agent-kit-starter-pack@claude-plugins-official' for Claude Code, and 'codex plugin marketplace add GoogleCloudPlatform/data-agent-kit-plugin' followed by 'codex plugin add dak@dak-marketplace' for Codex CLI.
After installation, sign in with your Google Cloud account and select the services you use. Data Agent Kit automatically enables required APIs, installs matching skills, and configures MCP servers—no manual setup files needed. The extension also brings a lightweight Google Cloud console view into your editor for browsing data and running queries.
Use MCP tools and skills to query and manage data
Once installed, your agent gains access to over 15 Google Data Cloud services through MCP tools. These let it inspect schemas, run queries, read job logs, and manage resources in your live environment using your IAM permissions. For example, you can ask your agent to forecast demand and check inventory, and it will use Knowledge Catalog to find trusted tables, run a BigQuery forecast, and check stock in AlloyDB—all within your editor.
Google-authored skills teach your agent data best practices. The bigtable-basics skill designs schemas around access patterns and flags hotspots before creation. The bigquery-graph-author skill maps tables to nodes and edges for property graphs, validates relationships against data, and shows a plan for approval. The federate-lakehouse-catalog skill connects your Iceberg Lakehouse to AWS Glue or Databricks Unity Catalog for cross-cloud queries without ingestion pipelines.
Skills also optimize workflows: gcp-pipeline-orchestration schedules dbt or Dataform models with notebooks as Airflow pipelines, and troubleshooting skills for Airflow and Spark trace failures and propose fixes. In the IDE, you can follow pipeline runs on a visual canvas and click 'Diagnose' to send logs to your agent for analysis.
Try a first prompt and explore next steps
To verify everything works, try a prompt like: 'What are this week’s fastest rising search terms in the US that weren’t in last week’s top 10? Use the BigQuery public Google Trends dataset.' Your agent will use the public dataset via MCP tools, apply relevant skills for query optimization, and return results directly in your editor or terminal.
For hands-on learning, explore the official documentation and product overview page. Browse, star, or contribute to the open-source skills on GitHub to adapt them to your team’s standards. The repository includes starter packs for various agents and examples of custom skill development.
Complete guided codelabs: try the 'Analytics with Data Agent Kit and Antigravity IDE' codelab for building analytics workflows, or the 'Fraud detection pipeline with Data Agent Kit and Antigravity IDE' codelab for data science pipelines. Companion blogs provide deeper context on agentic analytics and pipeline orchestration.
What to do next
Start by installing the IDE extension or CLI plugin for your agent, then run a simple query against a public dataset to confirm access. Use the GitHub repository to explore and customize skills for your team’s data practices. Finally, work through one of the codelabs to build a full analytics or pipeline workflow end-to-end.
FAQ
Do I need to pay extra for Data Agent Kit?
No, Data Agent Kit is free to use. You only pay for the underlying Google Cloud services your agent accesses, such as BigQuery queries or Bigtable reads.
Can I use Data Agent Kit with agents outside of VS Code?
Yes, it works as a plugin for Antigravity CLI, Claude Code, and Codex CLI, and is pre-installed in Cloud Shell and Cloud Workstations.
How does Data Agent Kit handle security and permissions?
The agent connects using your user identity or a service account you impersonate, so IAM, row-level security, and column-level security policies apply automatically. Admins can further restrict access using VPC Service Controls and Principal Access Boundary.
Source: Data Agent Kit is now GA: Bring Google Data Cloud to any coding agent (GCP).
