Skip to main content

Trading MCP — What You Can Do With It

These are the patterns we see most often when teams plug the Bitquery MCP server into Claude, Cursor, ChatGPT, or Claude Code. You don't write SQL — you ask in plain English, the agent does the rest. Each pattern below shows the kind of question that works and what comes back.

For end-to-end worked answers — including the live data table and a chart — see the examples section.

"What are the top 10 Solana tokens by USD volume in the last 24 hours? Skip wash-traded pools."

You get a clean, ranked list of the most-traded tokens for any chain — your daily watchlist generator. The agent automatically applies Bitquery's outlier filter so the noise is already gone.

See it live: Hottest Solana Tokens.

2. Trader Analytics and Realized PnL​

"Pull every trade for wallet 7xKX… in the last 7 days. Compute realized PnL per token in USD."

Wallet-centric breakdown — token by token, with bought, sold, and net flow in USD. Same idea works for any wallet on any supported chain. For a GraphQL equivalent see the Traders API.

See it live: Decode a Whale Wallet.

3. OHLC Charts for Any Pool or Token​

"Give me 1-minute OHLC for the WIF/USDC pool on Raydium for the last 6 hours."

Pre-aggregated candles at any interval — 1m, 5m, 1h, daily — for any pool, token, or currency. Ready to drop into TradingView or your bot.

See it live: Build an OHLC Candle Chart.

4. Market Cap and FDV Monitoring​

"Which Base tokens crossed $10M market cap in the last 24 hours?"

Every token aggregate row carries the latest market cap and fully diluted valuation. Background on the math: Supply fields.

5. Wash-Trade and Outlier Filtering​

"Show me 24h volume per chain, but only counting clean (non-wash-traded) flow."

Bitquery's price-index ranking is baked into every row. A simple "skip the noisy pools" instruction in your prompt is enough — no model to train, no list to maintain. Companion read: How to filter anomaly prices and the price-index algorithm.

6. New Token / Launch Monitoring​

"How many new tokens launched on Pump.fun in the last hour? How does that compare to the 24h average?"

Track launchpad activity in real time across Pump.fun, LetsBonk, FourMeme, Boop, Bags, Believe, Heaven, Goonfi, Trends.fun, Meteora Dynamic Bonding Curve, and more. Spot meta cooldowns and frenzy hours.

See it live: Pump.fun Launch Pulse. Per-launchpad GraphQL coverage: Pump.fun, LetsBonk, FourMeme, Meteora DBC.

7. Cross-Chain Market Overview​

"For each chain, show 24h DEX volume, number of trades, and number of unique traders."

A single sentence gets you a chain-by-chain comparison — no per-chain GraphQL juggling. Useful for spotting chain rotations and weighing where to deploy capital next.

See it live: Cross-Chain DEX Snapshot.

8. DEX Market Share Battles​

"Which Solana DEX is winning today's volume? Meteora vs Pumpswap vs Raydium vs Orca."

The agent ranks DEXs (Meteora, Pumpswap, Raydium, Uniswap, PancakeSwap, …) by 24h volume. The same pattern answers "which DEX is the cheapest for this pair?", "which DEX is the most retail-driven?", or "which DEX has the most active market-makers?"

See it live: Solana DEX Market Share.

9. Sniping and Copy-Trading Research​

"Find Solana wallets that bought any token in the first 60 seconds of its first Pumpfun trade in the last 24 hours, then sold within 10 minutes for a positive USD PnL."

The same data behind production bots — Solana sniper, Base sniper, Arbitrum sniper, copy-trading bot — is queryable conversationally. Perfect for prototyping signals before you commit to a Kafka stream.

10. Slippage and Liquidity Inspection​

"For pool 0xabc… on the last 100 trades, compute realized slippage relative to the volume-weighted price."

Every trade row carries price, USD amount, and side — enough to derive realised slippage and effective depth. Conceptual companion docs: Ethereum slippage, Base slippage, BSC slippage.

11. AI Agents and Trading Copilots​

The MCP is also the data layer for autonomous trading workflows:

  • In a chat assistant — Claude, Cursor, or ChatGPT answers ad-hoc data questions without context-switching to a separate IDE.
  • In a coding agent — Claude Code or Cursor generates ready-to-run analyses, charts, and alerts grounded in real on-chain data.
  • In a custom agent loop — pipe MCP tool calls into your own agent skill or trading bot. See the AI Agent on Solana data and AI Agent on Base data walkthroughs.

Best Practices for Prompting​

You don't need to know SQL or the schema — but a few prompt habits make the agent's answers dramatically better.

1. Be explicit about the time window​

The trading dataset is huge. Always tell the agent the window you care about: "in the last 24 hours", "yesterday vs the day before", "since 09:00 UTC today". Without it, the agent may scan more data than it needs to (slow) or pick a default that doesn't match what you wanted.

2. Name the chain (or "all chains")​

Solana, Ethereum, BNB Smart Chain, Base, Arbitrum, Optimism, Polygon, and Tron are all in the same dataset. "on Solana" or "across all chains" keeps the agent's filter clean.

3. Ask for clean (non-wash-traded) data when relevant​

Just say "skip wash-traded pools" or "only clean volume". The agent will apply Bitquery's outlier filter automatically. Use this for token discovery, market-cap rankings, and any time you want a "real" view of the market.

4. Pin the granularity for charts​

For OHLC and trend questions, name the bucket size: "1-minute candles", "hourly volume", "daily for the last 30 days". Otherwise the agent guesses.

5. Give it the address if you have one​

If you know a token contract, pool address, or wallet, paste it. Address-based lookups are the fastest the MCP can do, and they avoid symbol collisions (there are dozens of tokens called "PEPE").

6. Ask for the data shape you want​

"Give me a markdown table I can paste", "return JSON for my script", "format it for a Telegram message" — the agent will adapt. For chart-ready output: "return rows with timestamp, open, high, low, close, volume".

7. Iterate, don't restart​

Once you've built a good question, refine it instead of starting over: "same query, but for Base", "same chart, but only memecoins", "same wallet, but for the last 7 days". The agent keeps context.

8. Trust the read-only sandbox​

The MCP only allows reads. The agent cannot delete, insert, drop, or modify anything — even if you ask it to. Explore freely.

When MCP, When GraphQL, When Kafka?​

NeedBest fit
Conversational analysis, ad-hoc questions, agent loopsMCP (this server)
Application backend, predictable contract, subscriptions, mempoolGraphQL API and WebSocket subscriptions
Lowest-latency, highest-throughput streaming for production botsKafka streams and gRPC streams
Pre-built OHLC, market cap, token metadata over GraphQLCrypto Price API

The MCP and the GraphQL API read the same dataset, so anything you discover via MCP is reproducible in GraphQL or your production stream.