The complete guide to setting up the Cryptohopper Market Data MCP

The Cryptohopper Market Data MCP gives any AI agent — Claude, Codex, and half a dozen others — a direct line to live orderbooks, tickers, and candles across every major crypto exchange. Once it's connected, you can ask your agent questions like "Run a 1-hour TA on SOL" or "Where's the best liquidity for a $50k BTC buy right now?" and get a real answer, backed by real data.

---

This guide walks you through the full setup, end to end. It assumes you know what MCP is; if not, start with the setup overview. Otherwise, let's get you connected.

Before you start

You need three things:

1. A Cryptohopper account. The free Pioneer tier is enough to test the MCP — you get 6,000 calls per week, limited to Binance, Coinbase, and Kraken, real-time data only. Higher tiers unlock more exchanges, historical data, and higher quotas. See the tier comparison for the full breakdown.
2. A Cryptohopper MCP API key. This is a bearer token you'll paste into your MCP client's config. Full walkthrough in the setup overview.
3. An MCP-compatible client. Almost every modern AI tool now supports MCP. We've got dedicated guides for the most common ones below.

The MCP endpoint you'll always be pointing to is:

https://mcp-data.cryptohopper.com/mcp

Keep that URL handy — you'll paste it into every client.

Pick your client

Setup steps differ slightly per client. Pick yours, follow the dedicated guide, then come back here for the smoke-test. Each guide shows the exact JSON or command to paste, and where to paste it. None of them takes more than five minutes.

• Set up Cryptohopper MCP in Claude Code (terminal)
• Set up Cryptohopper MCP in the Claude desktop app
• Set up Cryptohopper MCP in Codex

The mental model — what happens when you connect

When you add the Cryptohopper MCP to your client, two things happen behind the scenes:

1. Tool discovery. The client connects to mcp-data.cryptohopper.com/mcp, authenticates with your bearer token, and asks: "what tools do you expose?" The server replies with a tidy list — get_ticker, get_orderbook, get_candles, get_market, list_markets, list_exchanges, list_quote_currencies, and get_usage. Each tool has a schema the model can read.
2. Live delegation. When you ask your model something that needs live data, it picks the right tool, fires the call, and uses the response to answer you. You never see the JSON unless you ask to.

If your client has a status or devtools panel, check it right after setup. You should see "Cryptohopper.com Market Data" listed as connected, with eight tools exposed.

The eight tools, at a glance

It's worth knowing what's actually on the menu before you start prompting — it shapes what you can ask for.

• get_ticker — last price, bid, ask, 24h high/low/volume, and 24h price change for a pair. The fastest way to answer "what's X trading at right now?" — ticker tool reference.
• get_candles — OHLCV candlestick data for a pair. Without arguments it returns the single latest cached candle; with limit + period it returns the latest N candles (max 1000); with start/end it returns an explicit historical range. Supported periods: 1m, 5m, 15m, 1h, 4h, 1d — candle (OHLCV) tool reference.
• get_orderbook — bid/ask snapshots as [price, volume] tuples. Latest snapshot by default (~150 levels live), or aggregated historical snapshots with start/end/period — orderbook tool reference.
• get_market — the full market definition for a single pair: price/amount precision, min/max order sizes, and maker/taker fees. Use this before computing valid lot sizes.
• list_markets — every trading pair available on an exchange, mapped to its market definition. Use this to discover valid base/quote combinations.
• list_exchanges — every supported spot exchange, with a last-health-check timestamp for each. A good first call to see what's reachable.
• list_quote_currencies — the quote currencies offered on an exchange and, for each, the base currencies paired against it. Handy for high-level discovery.
• get_usage — your current tier, per-minute usage vs. limit, weekly budget used vs. remaining, and when the budget resets.

Smoke-test in 60 seconds

Regardless of client, once it's connected, open a new chat and type:

"Using the Cryptohopper Market Data MCP, what's the current BTC/USDT ticker on Binance?"

A successful connection returns something like (illustrative values — yours will reflect the live market):

BTC/USDT on Binance — last ~$75,750, 24h range ~$75,310–$78,080,
24h volume ~16,800 BTC. Spread is tight. Data fresh as of 12:06 UTC.

If you get a real, current number back, you're done. Everything else from this point on is building on top.

Some common failure modes and what they mean:

• "Tool not found" — your model isn't connecting to the MCP. Usually a config error. Re-open the client's MCP panel and verify Cryptohopper is listed with its eight tools.
• "401 Unauthorized" — bad bearer token. Regenerate the key in your Cryptohopper account settings.
• "Exchange not supported" — you're on Pioneer or Explorer and asked for an exchange outside your tier's allow-list. Check the tier table.

We've collected the full error taxonomy in the error reference.

Your first real workflow

The fastest way to feel what MCP unlocks is to run a workflow that would have been fiddly to build by hand. Here are three starter prompts, each of which maps to a full tutorial if you want to go deeper.

1. On-demand technical analysis

"Pull 1-hour candles for ETH/USDT on Binance over the last 200 periods. Compute RSI(14), MACD(12,26,9), and the 20/50 EMAs. Tell me the trend and whether RSI is overbought or oversold."

You'll get a prose summary with indicator values. It's the sort of thing that would take ten minutes in TradingView, delivered in under ten seconds. Note the cap: a single get_candles call returns at most 1,000 candles, which is plenty for any standard indicator. Full recipe: run technical analysis on any token.

2. Top-movers scan

"From the pairs on Binance, pull recent daily candles for the majors and tell me which five had the largest percentage move in the last 24 hours. Summarise in a table."

Great way to see what's alive in the market without opening a browser. Full recipe: build a daily top-movers report.

3. Liquidity check before a trade

"I'm considering a $50k buy of BTC/USDT. What's the orderbook depth on Binance, Coinbase, and Kraken right now, and what's my expected slippage on each?"

The agent pulls orderbooks from all three venues and computes fill curves. Full recipe: estimate slippage before placing a trade.

If any of those worked, you've now built more AI-powered market analysis in ten minutes than most retail traders will have all year.

What the MCP does and does not do

It helps to be explicit here — it avoids surprises later.

The Cryptohopper Market Data MCP:

• Returns live orderbook, ticker, and candle data across supported exchanges.
• Lets you list available exchanges, trading pairs, and quote currencies.
• Returns your own weekly usage stats so you (or your agent) can stay within quota.

It does not:

• Execute trades. It's read-only. A dedicated trade-execution MCP is Coming soon.
• Schedule workflows. Use cron, Task Scheduler, or GitHub Actions — see how to schedule Cryptohopper MCP workflows.
• Store data between runs. If you want persistent memory, the agent needs to write to a file, a database, or a note.

Managing keys and usage

As you build more workflows, you'll want a clean story around keys and quota.

• One weekly quota, shared across all your keys. Rate limit and weekly total are bound to your user ID, not to an individual key. The weekly budget resets Friday at 00:00 UTC — worth noting if you batch heavy backtests near the end of the week.
• Tier-based key limits. Pioneer and Explorer can create one key; Adventurer three; Hero up to ten. Use multiple keys to separate different agents so you can rotate them independently — see run multiple agents with multiple API keys.
• Usage check, on demand. Any MCP client can call the usage endpoint. Just ask: "How much of my Cryptohopper quota have I used this week?" The full article: how to monitor your Cryptohopper MCP spend.

For a deeper dive on how calls are counted — especially the history-depth cost factor, which matters if you query long candle windows — see rate limits and cost factors explained.

Good defaults for new users

A few opinions from having watched hundreds of people get started.

• Start on Pioneer. If you can't confirm it does what you need on the free tier, an upgrade won't fix the problem.
• Pick one client first. Don't try to set it up in Claude Code and the Claude desktop app and Codex on day one. Get one working end-to-end, then replicate.
• Write down your prompts. As you iterate, save the prompts that actually produce useful output. They become your standard library. See write prompts that work with the Cryptohopper MCP.
• Watch your candle depth. Historical data costs more against your quota the further back you go, and a single call tops out at 1,000 candles. Most analyses only need the last 100–200 candles, not the full window. See rate limits explained for the exact multipliers.

Whichever direction you go, you've got the plumbing in place. The fun starts now.