Braian Arroyo Braconi
Back
Side project · 2026

Social Monitor

AI-assisted brand monitoring: it gathers what is published about a brand across the web and turns it into a weekly report.

Role
Design and development
Year
2026
Status
Working prototype
Stack
Next.js · Claude
Context

Listening for a brand without a heavy platform.

Social listening platforms tend to be expensive and complex to set up. I wanted a lightweight alternative: tell it which brand to follow and receive, every week, what actually matters.

How it works

From mention to report.

Collect

Connectors to news, social networks and forums search for each brand's terms.

Classify

Claude scores each mention: relevance, sentiment, topics, complaints and urgency.

Store

Mentions, brands and accounts live in Supabase.

Schedule

A Vercel cron collects every day.

Report

On Mondays it generates each brand's weekly report.

Decisions

Simple to start, easy to extend.

  • Value from day one

    Google News, GDELT, Mastodon and Hacker News work without credentials. Reddit, YouTube and Bluesky switch on as soon as their keys are added.

  • One connector, one interface

    Every source implements the same interface: adding a new one means writing one file and registering it.

  • Respecting limits

    The system honors each service's caps, such as one request every 5 seconds on GDELT or YouTube's daily quota.

  • Testing without infrastructure

    Scripts to run only collection, or collection plus AI, without a database.

Stack

Tools.

App
Next.js
Data and auth
Supabase
Classification
Claude API
Scheduled jobs
Vercel Cron
Sources
7 connectors
Output
Weekly report per brand
Next project WeeklyGoals

Let's talk.

Write to me