Gemini API Docs Preview Lands Inside Google AI Studio

On September 10, 2026, Google AI Studio shared the first preview of the Gemini API docs inside AI Studio — framed as docs built "from the ground up" for consolidation and convenience: docs right where you're building. Start at ai.studio/docs.

That lands in the same product surface explainx.ai has tracked through custom *.ai.studio URLs, agentic video understanding, and 3.7 Flash showcases in Studio and Antigravity: AI Studio as the place you prototype and look things up.

TL;DR

What Google actually announced

The official line is short and product-shaped, not a full docs rewrite:

today we're sharing the first preview of the Gemini API docs right in AI Studio — we built the experience from the ground up to give you better consolidation and convenience: docs right where you're building

Preview language matters. This is Google shipping an in-Studio docs experience, not claiming that every historical page on ai.google.dev has been retired. Treat ai.studio/docs as the new front door when you are already in Studio; keep the developer site for deep links, SDK references, and the coding-agents setup guide.

Why "docs where you build" is the real story

Community reaction zeroed in on the workflow win: documentation next to the prompt, key, and deploy flow kills context switching. Anyone who has held a Studio session in one window and three ai.google.dev tabs in another already knows the tax — copy a parameter name, lose the chat scroll position, forget which model ID the snippet assumed.

That same friction shows up whenever builders wire agents to docs instead of human browsers. explainx.ai's Markdown for Agents and llms.txt coverage is about serving machines clean text; this preview is the human-side twin — put the reference inside the build surface so you do not leave it.

Philipp Schmid's three tips builders should run today

Google DevRel engineer Philipp Schmid highlighted the agent-ready docs stack around the same preview. Three concrete moves:

1. Append .md to any docs URL

Need a page without chrome for an agent, a PR comment, or a notes vault? Append .md to the Gemini docs URL and pull Markdown instead of HTML. That matches the broader industry pattern of content negotiation for agents — same idea as serving Markdown to crawlers and assistants, documented on explainx.ai in the markdown-for-agents guide.

2. Wire the Gemini Docs MCP

npx add-mcp "https://gemini-api-docs-mcp.dev"

Google's Set up your coding assistant with Gemini MCP and Skills page documents a public Docs MCP at https://gemini-api-docs-mcp.dev. It exposes search_documentation so Claude Code, Cursor, Antigravity, Gemini CLI, and similar hosts can fetch current API patterns instead of inventing last year's SDK. For the protocol background, see explainx.ai's MCP complete guide and the skills + loops + MCP stack.

3. Install Gemini agent skills

npx skills add google-gemini/gemini-skills

For the Interactions API–focused skill specifically:

npx skills add google-gemini/gemini-skills --skill gemini-api-dev --global

Skills encode SDK and model hygiene (current model IDs, preferred patterns) into the agent. With Docs MCP connected, the skill prefers live doc search; without MCP it can fall back to llms.txt on ai.google.dev. That is the same portable skills pattern explained in What are agent skills?, Google's earlier google/skills pack, and the npx skills install flow. Browse more on explainx.ai's /skills and /mcp-servers directories.

"Ask Gemini" inside the docs

Another community beat on the preview: an "Ask Gemini" affordance in the docs UI — ask the documentation itself without opening a separate chat window. Combined with in-Studio hosting, that closes the loop Google is selling: read, ask, copy a snippet, jump back to Build without losing place.

Treat "Ask Gemini" as a convenience layer over the same docs corpus, not a substitute for verifying model IDs and pricing on the published pages — same discipline you use when an agent answers from MCP search.

How this fits the rest of the Gemini / AI Studio stack this month

September 2026 has already stacked several Studio-adjacent Gemini moves:

The throughline: Google is collapsing prompt → look up → ship into fewer surfaces. Custom URLs made deploys shareable; in-Studio docs make the reference local; MCP and skills make the same corpus available to coding agents outside the browser.

Honest limitations

• Preview, not a finished docs migration. Expect gaps, missing deep pages, or UI that still deep-links out to ai.google.dev for some topics.
• MCP and skills are separate installs. Opening ai.studio/docs does not automatically configure your local Claude Code or Cursor host — you still run add-mcp and skills add.
• .md URLs are a fetch tip, not a guarantee every historical path works. If a page 404s with .md, fall back to the HTML URL or MCP search.
• Training-cutoff agents still need the live tools. Without Docs MCP or a fresh Markdown paste, assistants will keep inventing deprecated generateContent-era patterns even while Studio shows Interactions API defaults.

What to do this week

1. Open ai.studio/docs from a logged-in AI Studio session and try a lookup you normally do on ai.google.dev.
2. Append .md to one docs URL and drop the Markdown into your agent or notes.
3. Run npx add-mcp "https://gemini-api-docs-mcp.dev" and ask your agent a current API question (Google suggests verifying with something like context caching).
4. Install google-gemini/gemini-skills (at least gemini-api-dev) and confirm /skills or your host's skills UI lists it.
5. If you care about token cost while agents re-read the same docs, revisit Gemini context caching for harnesses.

Related reading

• Google AI Studio custom URLs — free *.ai.studio subdomains
• Gemini agentic video understanding in Google AI Studio
• Gemini 3.7 Flash showcase across Antigravity and AI Studio
• Markdown for Agents — content negotiation for AI crawlers
• llms.txt specification guide
• What is MCP? Model Context Protocol guide
• What are agent skills? Complete guide
• Three AI skills for 2027 — skills, loops, MCP
• Google skills for Gemini and Google Cloud
• npx skills install guide

Official: ai.studio/docs · Gemini API docs · Coding agents: Gemini Docs MCP and Skills · Google AI Studio announcement

Preview availability, MCP endpoint, and skill package names accurate as of September 11, 2026, per Google AI Studio's September 10 announcement and ai.google.dev coding-agents documentation. Verify current install commands and docs URLs before wiring production agents.