Uploaded August 2026 | Updated September 2026, 2 weeks ago
In this episode of Hands On AI, Jeff Nelson walks through the Data Agent Kit， Google's unified kit of Agent Skills, MCP servers, and IDE integrations that lets your coding agent securely query and act on your Google Cloud data.
We take one real business question end to end: average order value at Cymbal Pets fell from ~$110 to ~$103 in January 2025. The orders live in BigQuery, the customer and pet profiles live in Cloud SQL, and the marketing campaign data is a JSON file in Cloud Storage.
We run the same workflow in the Antigravity IDE and in Claude Code, because the kit works anywhere， Antigravity, Cursor, any VS Code fork, Claude Code, Codex, Gemini CLI, and Antigravity CLI.
What you'll learn:
✅ MCP vs Agent Skills — the connector vs the guidebook (and when each fires)
✅ One-click MCP for BigQuery, Cloud SQL, AlloyDB, Spanner, and Knowledge Catalog
✅ Querying across BigQuery + Cloud SQL + Cloud Storage from a single chat
✅ Why transactional and analytical databases both exist (row store vs columnar)
✅ Installing the same kit as a plugin in Claude Code, Codex, and Gemini CLI
✅ Having an agent build, run, and test a dbt pipeline for you
✅ The one-to-many join bug that silently triples revenue — and the uniqueness test that catches it
✅ Zero-shot forecasting with BigQuery AI.FORECAST (TimesFM)
✅ Generating product copy and embeddings with ML.GENERATE_TEXT to power semantic search
📖 Chapters:
00:00 Intro — What is the Data Agent Kit?
00:46 - Skills, MCP tools, and IDE integrations
02:45 - Setting it up in the Antigravity IDE
04:14 - One-click MCP servers (BigQuery, Cloud SQL, Spanner, AlloyDB)
06:00 - Agent Skills: the Google Cloud guidebook
07:00 - MCP vs Skills — what's the actual difference?
09:08 - The case: "Why is our average order value dropping?"
11:19 - Where the data lives: BigQuery, Cloud SQL, Cloud Storage
14:19 - Transactional vs analytical databases (why you need both)
17:47 - Prompt 1: calculating monthly average order value
20:35 - The answer — and the January drop
21:51 - Drilling down: the B2B wholesale culprit
25:03 - Pulling customer profiles from Cloud SQL
27:18 - The same workflow in Claude Code (plugin install)
30:34 - Hunting the promo code: BIGORDER25
32:38 - The campaign JSON hiding in Cloud Storage
34:33 - The real answer — it was never a decline
35:45 - Productionizing the analysis with dbt
41:02 - The fan-out bug that would have tripled revenue
43:13 - A self-healing pipeline that passes its tests
46:01 - Exec summary + forecasting with AI.FORECAST (TimesFM)
50:01 - Generating a Jupyter notebook with charts
52:23 - Live demo: updating a real website from the agent
55:12 - Fixing semantic search with embeddings
57:22 - Who this is really for
58:19 - Wrap-up — try it on your own data
🔗 *Resource links:*
• Data Agent Kit → https://goo.gle/4zz5jM3
• Data Agent Kit — VS Code extension → https://goo.gle/45s8lUp
• Data Agent Kit — Claude Code plugin→ https://goo.gle/4xBZj2W
• Antigravity IDE → https://goo.gle/4bQox5n
• BigQuery → https://goo.gle/4xC3EDt
• Cloud SQL → https://goo.gle/4wRsbEy
• dbt → https://goo.gle/4gu4nAK
#DataAgentKit #BigQuery #AIAgents #GoogleCloud #MCP #dbt
Watch more Hands on AI → youtube.com/playlist?list=PLIivdWyY5sqKnJOvP89yF8t9mWuzMTcbM
🔔 Subscribe to Google Cloud Tech → https://goo.gle/GoogleCloudTech
Speakers: Annie Wang, Jeff Nelson
Products Mentioned: BigQuery, Data Agent Kit, Cloud SQL, Spanner, AlloyDB, Cloud Storage, Antigravity
In this episode of Hands On AI, Jeff Nelson walks through the Data Agent Kit， Google's unified kit of Agent Skills, MCP servers, and IDE integrations that lets your coding agent securely query and act on your Google Cloud data.
We take one real business question end to end: average order value at Cymbal Pets fell from ~$110 to ~$103 in January 2025. The orders live in BigQuery, the customer and pet profiles live in Cloud SQL, and the marketing campaign data is a JSON file in Cloud Storage.
We run the same workflow in the Antigravity IDE and in Claude Code, because the kit works anywhere， Antigravity, Cursor, any VS Code fork, Claude Code, Codex, Gemini CLI, and Antigravity CLI.
What you'll learn:
✅ MCP vs Agent Skills — the connector vs the guidebook (and when each fires)
✅ One-click MCP for BigQuery, Cloud SQL, AlloyDB, Spanner, and Knowledge Catalog
✅ Querying across BigQuery + Cloud SQL + Cloud Storage from a single chat
✅ Why transactional and analytical databases both exist (row store vs columnar)
✅ Installing the same kit as a plugin in Claude Code, Codex, and Gemini CLI
✅ Having an agent build, run, and test a dbt pipeline for you
✅ The one-to-many join bug that silently triples revenue — and the uniqueness test that catches it
✅ Zero-shot forecasting with BigQuery AI.FORECAST (TimesFM)
✅ Generating product copy and embeddings with ML.GENERATE_TEXT to power semantic search
📖 Chapters:
00:00 Intro — What is the Data Agent Kit?
00:46 - Skills, MCP tools, and IDE integrations
02:45 - Setting it up in the Antigravity IDE
04:14 - One-click MCP servers (BigQuery, Cloud SQL, Spanner, AlloyDB)
06:00 - Agent Skills: the Google Cloud guidebook
07:00 - MCP vs Skills — what's the actual difference?
09:08 - The case: "Why is our average order value dropping?"
11:19 - Where the data lives: BigQuery, Cloud SQL, Cloud Storage
14:19 - Transactional vs analytical databases (why you need both)
17:47 - Prompt 1: calculating monthly average order value
20:35 - The answer — and the January drop
21:51 - Drilling down: the B2B wholesale culprit
25:03 - Pulling customer profiles from Cloud SQL
27:18 - The same workflow in Claude Code (plugin install)
30:34 - Hunting the promo code: BIGORDER25
32:38 - The campaign JSON hiding in Cloud Storage
34:33 - The real answer — it was never a decline
35:45 - Productionizing the analysis with dbt
41:02 - The fan-out bug that would have tripled revenue
43:13 - A self-healing pipeline that passes its tests
46:01 - Exec summary + forecasting with AI.FORECAST (TimesFM)
50:01 - Generating a Jupyter notebook with charts
52:23 - Live demo: updating a real website from the agent
55:12 - Fixing semantic search with embeddings
57:22 - Who this is really for
58:19 - Wrap-up — try it on your own data
🔗 *Resource links:*
• Data Agent Kit → https://goo.gle/4zz5jM3
• Data Agent Kit — VS Code extension → https://goo.gle/45s8lUp
• Data Agent Kit — Claude Code plugin→ https://goo.gle/4xBZj2W
• Antigravity IDE → https://goo.gle/4bQox5n
• BigQuery → https://goo.gle/4xC3EDt
• Cloud SQL → https://goo.gle/4wRsbEy
• dbt → https://goo.gle/4gu4nAK
#DataAgentKit #BigQuery #AIAgents #GoogleCloud #MCP #dbt
Watch more Hands on AI → youtube.com/playlist?list=PLIivdWyY5sqKnJOvP89yF8t9mWuzMTcbM
🔔 Subscribe to Google Cloud Tech → https://goo.gle/GoogleCloudTech
Speakers: Annie Wang, Jeff Nelson
Products Mentioned: BigQuery, Data Agent Kit, Cloud SQL, Spanner, AlloyDB, Cloud Storage, Antigravity
