CT sentiment intelligence for Solana. Scans Crypto Twitter every 15 minutes. Scores narratives. Tells you what the market is thinking before the price moves.
Crypto Twitter moves before the chart does. But 99% of it is noise — shills, bots, and emotional reactions. The 1% that matters is specific, credible, and high-engagement.
Echo fetches recent tweets about your tracked tokens, aggregates mentions by engagement weight, and passes the data to a Claude agent trained to separate signal from noise. The output is a ranked sentiment leaderboard with confidence scores and one-line action hints.
FETCH → AGGREGATE → WEIGHT → REASON → SIGNAL → RANK
Works with or without a Twitter API key — falls back to realistic mock data for testing.
┌────────────────────────────────────────────────────┐
│ Twitter Feed Layer │
│ v2 Search API · Engagement scoring │
│ Fallback: mock data (no API key needed) │
└───────────────────────┬────────────────────────────┘
▼
┌────────────────────────────────────────────────────┐
│ Mention Aggregator │
│ Count · Engagement weight · Influencer flag │
│ Momentum detection (vs previous cycle) │
└───────────────────────┬────────────────────────────┘
▼
┌────────────────────────────────────────────────────┐
│ Claude Sentiment Agent │
│ get_tweet_sample → get_mention_stats │
│ → get_narrative_keywords → submit_signal │
└───────────────────────┬────────────────────────────┘
▼
┌────────────────────────────────────────────────────┐
│ Ranker + Reporter │
│ 24h rolling history · Leaderboard │
│ Scan report with top narratives │
└────────────────────────────────────────────────────┘
| Signal | Meaning | Trading Implication |
|---|---|---|
| bullish | Credible positive consensus | Consider entry |
| bearish | Credible negative consensus | Avoid or reduce |
| hype | High engagement, possibly unsustainable | Take profit zone |
| fud | Fear/uncertainty spreading | Check if substantiated |
| neutral | Mixed or low-signal | Wait for clarity |
- Accounts 10k+ followers:
3×engagement weight - Retweets:
3×likes weight - Specific claims ("accumulating", "whale"): quality boost
- Generic posts ("LFG", "gm"): filtered out
git clone https://github.com/EchoSentiment/EchoSentiment
cd EchoSentiment && bun install
cp .env.example .env
bun run devNo Twitter API key needed — mock data is built in.
ANTHROPIC_API_KEY=sk-ant-...
TWITTER_BEARER_TOKEN=... # optional
TRACKED_TOKENS=SOL,JTO,BONK,WIF,PENGU,JUP,RAY
SCAN_INTERVAL_MS=900000 # 15 min
MIN_ENGAGEMENT_SCORE=50
MIN_INFLUENCER_FOLLOWERS=10000MIT
read the room. front-run the narrative.