Crypto Research Agent

The Problem

A crypto trader had a proven research strategy — but executing it meant hours of scrolling Twitter every day, tracking which tokens were gaining buzz across dozens of accounts. The strategy worked, but the process didn't scale.

What We Built

We took the time to understand his exact research strategy — then built an AI agent that thinks the way he does. The system monitors the same 49 Twitter accounts he was already following, custom-tailored to his workflow:

Monitors the 49 Twitter accounts he already follows — custom to his watchlist

Calculates momentum by comparing current mentions against a 7-day baseline

Uses AI to analyze sentiment and generate summaries using his research lens

Sends real-time Telegram alerts when momentum signals spike

How It Works

Three automated steps run daily — no manual intervention needed. Scan Twitter for mentions, analyze momentum and sentiment with AI, and deliver Telegram alerts when thresholds are hit.

Momentum Scoring

Compares today's mentions against a 7-day baseline. If a token normally gets 5 mentions/day but suddenly gets 17 weighted mentions, that's a 3.4x momentum spike — and an alert fires.

VIP Weighting

Not all mentions are equal. Top alpha callers are weighted 3x, so one VIP mention carries the same signal as three regular ones.

Smart Filtering

BTC, ETH, SOL and ~20 other major coins are excluded. The agent only tracks smaller tokens with breakout potential.

Alert Cooldown

Once a token triggers an alert, it won't alert again for 24 hours — unless momentum doubles. No spam, just signal.

Cold Start Logic

New tokens the agent hasn't seen before get a conservative baseline so they don't trigger false alerts on day one.

Results

Accounts Monitored

Daily

Automated Scans

98%

Less Time Researching

Real-time

Telegram Delivery

Tech Stack

Next.jsSupabaseVercel AI GatewaySocialData APITelegram Bot API