Gemini-API is a reverse-engineered Python wrapper for the Google Gemini web app, giving developers programmatic access to Gemini’s consumer-facing features through an elegant, asynchronous interface. It mirrors the style of Google’s official Generative AI API while unlocking capabilities of the web experience that the official API does not expose.
The library is open source on GitHub with about 3,620 stars and active development in 2026, and it is published on PyPI as the gemini-webapi package. It handles persistent cookies with automatic background refresh, making it suitable for always-on services. Beyond text chat, it supports native image generation and editing, video and audio generation, the full Deep Research workflow with plan creation and status polling, custom system prompts through Gemini Gems, and Gemini extensions such as YouTube and Gmail. Outputs are classified into text, thoughts, images, videos and audio, and a streaming mode plus a standalone CLI are included. Because it is reverse-engineered rather than official, users should expect occasional breakage when Google changes the web app. It is one of the most starred unofficial Gemini clients and is updated frequently as Google’s web interface evolves, with releases tracked on PyPI so downstream projects can pin a working version.
Gemini-API is free and open source under the AGPL-3.0 license. It works with a regular Gemini web account, though advanced models and generation features depend on your Google plan.
Python developers and hobbyists who want scriptable access to Gemini web features like Deep Research and media generation before they exist in the official API.