What is chrome-ai-tools?

chrome-ai-tools lets you call Chrome's built in AI features, which run on a model called Gemini Nano, from a command line, from Node.js, or from Python, instead of only from inside a browser page. Chrome's AI APIs are normally locked to running inside browser pages, so this project bridges that gap. A small Python server manages a queue of prompts, and a page you open once in Chrome picks up those prompts and actually calls the AI directly, then sends the result back to the server. You install it with npm, then start the included Python server, which listens on a local address, and open that address once in a Chrome browser tab, leaving it open in the background. After that, you can send prompts using a command line tool called chrome-ai, or by importing helper functions into a Node.js or Python project, or even by sending plain HTTP requests with a tool like curl. Four kinds of tasks are supported: general prompting with an optional system instruction, summarizing a block of text, translating text between two languages, and writing new text such as a poem, given an optional context. Each of these can be piped text from a file, called from code, or requested with a single command line call. No browser extensions, no separate automated browser tooling, and no API keys are required, only a recent desktop version of Chrome and Python 3.9 or newer. The project is described as alpha status: it is not yet published to npm, the interfaces may still change, and some edge cases are not yet handled, but it works for prototyping through a local install. It also ships as an installable skill for a coding agent called Pi.