Skip to content

Brian24NX/proxy-ai

Repository files navigation

Proxy AI — Your AI Meeting Assistant

Live app: https://devfest-washu-2026.vercel.app GitHub: https://github.com/Brian24NX/proxy-ai


What Is Proxy AI?

Meetings are full of decisions, action items, and follow-ups — and keeping track of all of it is exhausting. Proxy AI is a web app that reads a meeting transcript or description you provide and turns it into a clean, organized summary in seconds.

Just paste in what happened during your meeting (or describe it in your own words), and Proxy AI will hand it off to a team of four specialized AI agents that work together to produce a polished, ready-to-share summary — complete with key decisions, action items with owners and deadlines, and discussion highlights.

No account required. No downloads. Open the link, paste your meeting, and get results.


How It Works

When you submit a meeting, your input passes through four AI agents in sequence. Each one has a specific job, and each one builds on the work of the one before it.

Your Meeting Input
        |
        v
+-------------------+
|    CEO Agent      |  Understands what you actually need.
|                   |  Sets the goal, priorities, and boundaries
|                   |  for the rest of the team.
+-------------------+
        |
        v
+-------------------+
|   Product Agent   |  Translates the CEO's direction into a
|                   |  clear plan — defining exactly what the
|                   |  final output should look like.
+-------------------+
        |
        v
+-------------------+
| Operator Agent    |  The one who does the actual work.
|   (Proxy)         |  Reads the meeting content and produces
|                   |  a structured analysis with summaries,
|                   |  decisions, and action items.
+-------------------+
        |
        v
+-------------------+
|   Critic Agent    |  Reviews everything before it reaches
|                   |  you. Catches gaps, unclear items, and
|                   |  missing context — then delivers a final
|                   |  polished result.
+-------------------+
        |
        v
  Your Final Output

Each agent runs one at a time. You can watch the progress live on the page as each step completes.


Getting Started

You have two options: use the live app online (no setup needed), or run it on your own computer.

Option 1: Just Use the Live App (Recommended)

No setup required. Visit:

https://devfest-washu-2026.vercel.app

That's it. Skip to the "How to Use" section below.


Option 2: Run It on Your Own Computer

This option is for people who want to run the project locally — for example, to customize it or explore the code. These instructions assume you have never used a terminal before.

Step 1: Check if you have Node.js installed

Node.js is the software that runs this project on your computer. To check if you already have it:

  1. On a Mac, open the Terminal app. You can find it by pressing Command + Space and typing "Terminal."
  2. On Windows, open Command Prompt. Press the Windows key, type "cmd," and press Enter.
  3. Type the following and press Enter:
node --version

If you see something like v20.0.0 or higher, you already have Node.js. If you see an error, go to https://nodejs.org and click the big green "LTS" download button, then run the installer.

Step 2: Download the project

Go to https://github.com/Brian24NX/proxy-ai and click the green Code button, then click Download ZIP. Once downloaded, unzip the file. You should have a folder called proxy-ai-main (or similar).

Alternatively, if you have Git installed, you can run this in your terminal:

git clone https://github.com/Brian24NX/proxy-ai.git

Step 3: Open the project folder in your terminal

In your terminal, you need to navigate into the project folder. Type cd (with a space after it), then drag the project folder from your file manager into the terminal window — it will automatically fill in the path. Then press Enter.

It will look something like this:

cd /Users/yourname/Downloads/proxy-ai-main

Step 4: Get a Google Gemini API key

Proxy AI uses Google's Gemini AI model to power its agents. You need a free API key to use it.

  1. Go to https://aistudio.google.com/app/apikey
  2. Sign in with a Google account
  3. Click Create API Key
  4. Copy the key — it will look like a long string of letters and numbers

Step 5: Add your API key to the project

In the project folder, create a new file called .env.local. You can do this in your terminal:

On Mac/Linux:

touch .env.local

On Windows:

type nul > .env.local

Then open .env.local in any text editor (Notepad works fine on Windows, TextEdit on Mac). Add this line, replacing your-key-here with the actual key you copied:

GEMINI_API_KEY=your-key-here

Save the file.

Step 6: Install the project's dependencies

In your terminal, type this and press Enter:

npm install

This downloads everything the project needs to run. It may take a minute or two. You will see a lot of text — that is normal.

Step 7: Start the app

Type this and press Enter:

npm run dev

Once you see a message saying the server is ready, open your web browser and go to:

http://localhost:3000

The app is now running on your computer. To stop it later, go back to the terminal and press Control + C.


How to Use Proxy AI

Step 1: Open the app

Go to https://devfest-washu-2026.vercel.app (or http://localhost:3000 if running locally).

Step 2: Enter your meeting content

You will see a text area on the page. Paste or type one of the following:

  • A transcript of your meeting (copied from Zoom, Google Meet, Microsoft Teams, Otter.ai, etc.)
  • A rough description of what happened — even informal notes work
  • A summary you wrote yourself that you want cleaned up

There is no strict format required. Write naturally.

Example of what you might type:

"We had a meeting today with the marketing and engineering teams. We decided to delay the product launch to May 15th. John is responsible for updating the website, and Sarah will send out the email announcement by Friday. We also discussed expanding to a new market but decided to revisit that next quarter."

Step 3: Submit

Click the button to run the analysis.

Step 4: Watch the agents work

You will see four steps appear on the screen, each lighting up as an agent finishes its job. The full process typically takes 20 to 60 seconds, depending on the length of your input and current server load. Be patient — the agents are doing real work in the background.

Step 5: Read your results

When all four agents are done, a final polished output will appear. It will include:

  • A concise summary of the meeting
  • Key decisions that were made
  • An action item table with owners, tasks, and deadlines
  • Discussion highlights and suggested follow-ups

You can copy this output and paste it into an email, a Slack message, a Notion page, or wherever your team keeps records.


Example Use Cases

Proxy AI works best when there is real meeting content to process. Here are some situations where it shines:

Team standups and check-ins Paste in a brief description of your daily or weekly sync. Proxy AI will pull out blockers, updates, and next steps.

Project kickoff meetings Describe the goals, stakeholders, and initial decisions discussed. Get back a clean brief you can share with the whole team.

Client or sales calls Paste in notes or a transcript. Proxy AI organizes what was promised, what was agreed to, and what needs follow-up.

Retrospectives Describe what went well, what did not, and what the team wants to change. Get a structured retrospective summary ready to share.

Planning sessions Describe the features, timelines, or priorities discussed. Get back a structured summary with clear ownership.

General team meetings Even messy, free-form meetings become organized documents. If it happened in a meeting, Proxy AI can help make sense of it.


Frequently Asked Questions

Is it free?

Yes. The live app at https://devfest-washu-2026.vercel.app is completely free to use. No account, no credit card, no subscription. If you run it locally, you will use your own Google Gemini API key — Gemini offers a free tier that is sufficient for personal use.

Is my data safe?

Your meeting content is sent to Google's Gemini API for processing. It is not stored by Proxy AI itself. However, you should be mindful of sharing sensitive or confidential meeting content through any AI tool. Avoid pasting information that includes passwords, personal health information, or legally sensitive material. Review Google's privacy policy if you need to understand exactly how Gemini handles input data.

What kind of meetings can I use this for?

Any kind of meeting. Team standups, project planning sessions, one-on-ones, client calls, retrospectives, board meetings, brainstorming sessions — Proxy AI is not limited to a specific type. As long as you can describe or paste what was discussed, it can work with it. It performs best when given more detail rather than less.

How accurate is it?

Proxy AI is as accurate as the information you give it. If your transcript or description is complete, the output will reflect that well. If your input is vague or missing key details, the output may also be incomplete — and the Critic Agent will flag those gaps. AI models can occasionally misinterpret phrasing or make minor errors, so always review the output before sharing it with your team. It is a tool to assist you, not replace your judgment.

Should I let other meeting participants know that Proxy AI is being used?

Yes, we strongly recommend it. Transparency matters — meeting participants should know when their words are being recorded, transcribed, or processed by an AI tool. Before using Proxy AI, let your team or meeting attendees know that you plan to run the meeting notes through an AI assistant. This builds trust, respects everyone's privacy, and in some workplaces or regions may be legally required. A simple heads-up like "I'm going to run our notes through an AI summarizer to share with the team" is enough.


Tech Stack

For those who are curious about what powers Proxy AI:

Technology What It Does
Next.js 16 The web framework the app is built on
React 19 Powers the interactive user interface
TypeScript Adds type safety to the codebase
Tailwind CSS Handles the visual styling and layout
Framer Motion Animates the agent pipeline steps
Google Gemini 2.5 Flash The AI model that powers all four agents
Vercel Hosts and deploys the live app

Credits

Proxy AI was built for DevFest WashU 2026, the annual hackathon hosted at Washington University in St. Louis.

Built by Brian Zhou, Alice Wang, and Linda Wang — GitHub: https://github.com/Brian24NX

Special thanks to the DevFest WashU organizers and everyone who made the event possible.

About

Proxy AI - AI-powered meeting assistant with multi-agent system. DevFest WashU 2026.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors