Airia

Inspiration

Email remains one of the most important tools for communication in professional environments, but managing inboxes can quickly become overwhelming. Professionals spend a significant amount of time reading emails, deciding what action is needed, drafting responses, and tracking tasks mentioned in messages. I wanted to explore how AI agents could be embedded directly into existing workflows to reduce this cognitive load. The idea behind Airia was simple: instead of forcing users to leave their email client to use an AI assistant, bring the AI directly into the inbox. This led to the concept of a browser-based AI agent that lives inside Gmail and assists users with understanding emails, drafting responses, and extracting actionable tasks.

What it does

Airia is an AI-powered email assistant implemented as a Chrome extension that integrates directly into Gmail. It provides three main capabilities: Email Analysis – The system analyzes the email content and helps users understand the message quickly. AI Reply Drafting – Airia generates professional draft replies to emails directly inside the Gmail interface. Task Extraction – The system identifies actionable items within the email and presents them as tasks. All of these actions happen within the Gmail interface, meaning users do not need to switch tools or open another application.

How we built it

The project consists of two main components:

Chrome Extension (Frontend) A custom Chrome extension injects an Airia panel directly into the Gmail interface. The extension uses DOM parsing to capture email content such as the sender, subject, and message body.
FastAPI Backend (AI Agent Layer) The extension communicates with a Python backend built using FastAPI. The backend acts as the AI agent layer responsible for: Email analysis Draft reply generation Task extraction The system processes email content and returns structured outputs that are displayed directly within Gmail. Architecture Overview

Gmail Interface ↓ Chrome Extension (Airia UI) ↓ FastAPI Backend (AI Agent) ↓ Email Analysis / Reply Generation / Task Extraction

Challenges we ran into

One of the main challenges was integrating the AI assistant directly into Gmail. Gmail is a dynamic single-page application, which means the page does not fully reload when emails change. This required carefully detecting DOM updates to ensure the extension correctly captured the current email content. Another challenge was designing the interaction between the extension and the backend so that responses were fast and seamless, maintaining a smooth user experience inside the Gmail interface. Another challenge is the use of the model I used will building it. I used Gemini 2.5-flash model which is somehow outdated, I don't have money to buy the latest model from chatgpt or anthropic

Accomplishments that we're proud of

One of our biggest accomplishments was successfully embedding an AI agent directly into the Gmail interface through a Chrome extension. Instead of creating a separate application, we built Airia to work inside an existing workflow, allowing users to analyze emails, generate replies, and extract tasks without leaving their inbox. We are also proud of building a full end-to-end system as a solo project. The solution combines a browser extension for real-time interaction with Gmail and a Python FastAPI backend that powers the AI agent responsible for email understanding, reply generation, and task extraction.

What we learned

This project highlighted the power of embedding AI agents directly into existing tools rather than building standalone applications. By integrating AI into familiar workflows like email, productivity tools can become significantly more efficient. It also reinforced the importance of combining frontend browser technologies with backend AI services to create real-world AI-powered systems.

What's next for Airia

Future improvements could include: Integration with additional platforms such as Outlook and Slack More advanced email classification and prioritization Calendar and task management integrations Multi-agent workflows for handling complex communication scenarios Airia demonstrates how AI agents can become a seamless part of everyday productivity tools.

Built With

css
fastapi
html
javascript
langchain
python

Submitted to

Airia AI Agents Hackathon

Created by

I designed and built the complete Airia platform end-to-end — from the AI agent pipeline to the Gmail Chrome extension. I architected the FastAPI backend, integrated Google Gemini 2.5 Flash as the core AI engine, and implemented a two-agent workflow where Agent 1 analyses and classifies emails while Agent 2 generates contextual draft replies. I built the Chrome extension that injects Airia directly into Gmail with real-time triage labels, one-click reply drafting, auto task extraction, and a privacy guardrail that detects and blocks generation on sensitive content. I also built the live monitoring dashboard and the clean HTML/JS frontend. This was my first time building a browser extension, and I learned a tremendous amount about Chrome extension APIs, content script injection, and shipping a full-stack AI product under time pressure.

Mustapha saheed

Updates

Mustapha saheed started this project — Mar 09, 2026 10:42 AM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.