Imagine you’re writing a LinkedIn post. You want to correct it, improve the styling, or even translate it to another language. The usual workflow is to copy the text into ChatGPT, Gemini, or any other LLM, ask for changes, and then paste the improved version back. It works, but it’s time-consuming and a bit clunky.

Product definition

I wanted to make this workflow seamless: no tab switching, no extra copy/paste.

The simplest and most natural solution was a browser extension that works on any website and can interact with text input fields. You select the input you want to improve and ask the assistant to make the changes without leaving the page.

The extension adds a button next to text inputs. Click it and a chat modal pops up: you can ask the assistant to fix grammar, improve style, translate, etc. The assistant then updates the field content according to your instructions.

I also wanted users to customize the assistant’s tone per website. For example, what you want on LinkedIn is usually different from what you want on a more informal website like Twitter. And if you give the assistant a bit of context (like your basic personal info), it can write in a more personal and consistent voice.

To achieve this, Sireno supports skills. Skills are small instruction files that tell the assistant how to behave. For example, a LinkedIn skill can ask for a more formal style and include a few details about you. Skills are only active on the domains you choose, so you can keep a formal assistant on LinkedIn and a more informal one on Twitter (or provide different context depending on the site).

And why stop at a single input field? In the side panel, you can select which fields on the page you want to interact with, configure how much page context the assistant can access, and fill multiple fields at once. For example, when you’re filling out a sign-up form, you can use a personal-data skill to fill everything in one go.

How to use it

You can install the extension from the Chrome Web Store. Once installed, you will see a new extension button in the extension bar. You can click it to open the side panel, where you can configure the AI provider in the settings tab. This requires an API key from any AI provider you want to use, which has a cost—not too much—but if you don’t have one, I plan to release an update in the coming weeks with support for Groq (not to be confused with Grok), which offers a free tier that can be used with this extension.

You can also manage your skills in the skills tab, check the fields available on the current page in the fields tab, and finally interact with the assistant in the chat tab.

I recommend reviewing the default skills and updating the one with your basic info (name, job, etc.). This helps the assistant write more personalized content. You can also create new skills for specific websites or use cases.

From there, you’re ready to use everything on any website. Click the button next to any text input field, or use the side panel chat to start interacting with the assistant.

Technical details

I built this project as an experiment with OpenCode as a vibe-coding tool. It helped me iterate quickly and ship an MVP fast, but I still had to get my hands dirty on the tricky parts that LLMs tend to miss: positioning the icon reliably, handling iframes, Shadow DOM, and so on. (I’ll talk about my experience with OpenCode in another post.)

This extension is open source on GitHub, so you can check the code and contribute if you want: https://github.com/sergiocarracedo/sireno-assistant

If you like it and find any issue or have any suggestions, please let me know in the GitHub repo issues

This is a side project that I built in my free time just to enjoy doing things, but it is an early MVP, so it may have some rough edges and bugs, but I hope to improve it over time based on user feedback and contributions.