Prerequisites
Cryptohopper MCP configured in an MCP client — see the setup overview.
Explorer tier or higher for candle-based trend analysis. Ticker-only watchlists run on the free Pioneer tier — see subscription tiers.
A list of 10–20 tokens you actually care about. If you haven't checked a coin's price in a month, it doesn't belong on the list.
Setup steps
Define your watchlist
Keep it tight — bigger watchlists get skimmed; a focused list gets read.
Issue the core digest prompt
Using the Cryptohopper MCP, build a morning digest for this watchlist:
BTC, ETH, SOL, AVAX, ARB, OP, SEI, LINK, AAVE, UNI — all on Binance,
paired against USDT.
For each token:
- Pull the current ticker (last price, 24h change %, 24h volume).
- Pull the last 50 4h candles.
- Identify the main trend (up, down, ranging) from the candles.
- Note whether 24h volume looks elevated vs. the recent 4h-candle
volume average.
Output as a markdown table with columns: token, price, 24h %, volume
state, 4h trend, one-line note.
Above the table, write a one-paragraph "what happened overnight"
summary covering any material changes from yesterday.Run and refine
The first output is typically 80% right and 20% generic. Make the "one-line note" more specific: "mention RSI, nearest support/resistance, or recent breakout/breakdown if any."
Parameterise the watchlist
Store the list at the top of a reusable prompt so you edit it in one place:
WATCHLIST = [BTC, ETH, SOL, AVAX, ARB, OP, SEI, LINK, AAVE, UNI]
EXCHANGE = Binance QUOTE = USDT
Add an overnight changes section
Two options: have the agent read the previous day's saved output and compute the delta, or have it pull daily candles and compare the last two closes — works without saving state.
Automate on a daily schedule
Once per weekday morning is the typical cadence. See how to schedule Cryptohopper MCP workflows.
Deliver to where you read
Telegram for a quick phone glance, email for longer digests, Notion or Google Doc for historical archive, local markdown file if you prefer to pull rather than push. See how to send MCP reports to Telegram, Discord, or email.
Sample output
Morning Digest — 24 april 2026
Overnight: Broad risk-on move across majors. ETH led with +2.8%; SOL and ARB followed. LINK and UNI lagged, both sideways on the day.
Token | Price | 24h % | Volume | 4h trend | Note |
BTC | 66,120 | +1.1% | Normal | Up | Holding above 65,800 support. |
ETH | 3,240 | +2.8% | Elevated | Up | Breaking out of 3-day range. |
SOL | 148.50 | +1.9% | Normal | Up | RSI 62, momentum intact. |
AVAX | 34.80 | +0.6% | Normal | Ranging | Tight consolidation, 34–35. |
ARB | 0.9720 | +2.1% | Elevated | Up | 2nd day of volume pickup. |
Cost profile
Action | Explorer | Hero |
10 tickers | 10 units | 10 units |
10 × 50 4h candles | ~50 units | 10 units |
Per run | ~60 units | ~20 units |
Weekly (5 runs) | ~300 units | ~100 units |
Easily within Explorer's 30,000 calls/week. Ticker-only watchlists run well within Pioneer's 6,000 calls/week. See rate limits explained.
Troubleshooting
The digest is technically correct but feels bland
The agent is giving generic trend labels. Require specific numeric observations: "mention at least one concrete number per token (price level, RSI value, % move over N days)" — not vague adjectives like "healthy" or "weak".
Half the tokens on the watchlist are ones you no longer follow
Refresh it quarterly. A dashboard you keep reading beats a comprehensive one you skip.
Volume state says "normal" for everything
The comparison baseline is too broad. Prompt more precisely: "compare current 24h volume against the median of the last 7 daily candle volumes; flag as elevated if above 2× median."
The digest runs but nothing arrives in Telegram
The most common cause is a silent failure in the delivery step — the MCP part worked but the Telegram post failed. Separate the two steps so failures surface independently. See how to schedule Cryptohopper MCP workflows.
You want to track changes over time
Append every morning's digest to a running markdown file or Google Sheet. Over weeks, you get a useful record of how each token moved — and how accurate the AI's one-line notes were.
You want per-token sparklines or charts
The MCP returns numeric data; rendering charts is outside its scope. Run the output through a charting library (matplotlib, Plotly) or deliver to Notion where embed blocks work. Most daily-digest readers prefer text-only anyway.