Google AI Studio’s web UI, dressed up as three major APIs

What it does

An adapter that sits between your HTTP client and Google AI Studio Build. It drives the AI Studio web UI via browser automation and reshapes the results to look like OpenAI, Gemini, or Anthropic API responses. Chat completions, embeddings, image generation, token counting, and tool calls are supported across all three facade formats.

The interesting bit

The real trick is multi-account rotation. It keeps several Google accounts logged in simultaneously and switches between them without re-authentication, turning a consumer web app into a rotatable backend. A built-in web console even exposes a VNC session so you can log into accounts visually when running inside Docker.

Key highlights

• Speaks OpenAI, Gemini, and Anthropic API formats from a single endpoint, forwarding everything through the Gemini-native interface internally.
• Supports true streaming, fake streaming, and non-streaming responses for chat endpoints.
• Handles tool calls across all three API dialects.
• Includes a web dashboard for account management, auth file upload and download, and VNC-based login.
• Persists usage statistics to a local JSONL file.

Caveats

• VNC login for adding accounts is only available under Docker; direct installs require a local setup script.
• Authentication cannot be injected via environment variables—you must supply auth files or log in through the browser.
• The OpenAI-compatible Responses API facade does not support image generation.

Verdict

Useful if you want to consume AI Studio models through standard API contracts without managing separate provider keys. Less appealing if you require a fully programmatic authentication flow, since the tool currently demands browser-based login and auth files rather than environment variables.