How to Use AI to Analyse On-Chain Crypto Data: ChatGPT, Claude, DeepSeek, Glassnode and Dune Workflow

The definitive 2026 guide: use Claude, ChatGPT, DeepSeek, Gemini and Grok to analyse Glassnode, CryptoQuant, Dune, Nansen data. 50 copy-paste prompts for every trader level. 

SUMMARY: AI language models can analyse on-chain crypto data without the trader needing any coding skills. The workflow has four steps: export or copy data from an on-chain analytics platform (Glassnode, CryptoQuant, Dune Analytics, Nansen, CoinGlass, DeFiLlama, or Etherscan), paste it into an AI model with a precisely structured prompt, request specific analytical output (trend interpretation, signal generation, scenario modelling, or trading plan construction), and act on the AI’s synthesis alongside your own judgement. Claude Sonnet 4.6 excels at structured multi-signal synthesis, long-context document analysis, and constructing rigorous trading frameworks. ChatGPT (GPT-5.4) excels at code generation, TradingView Pine Script writing, and news-to-on-chain correlation. DeepSeek V4 is the best free option for mathematical analysis and quantitative reasoning at near-zero cost. Gemini 3.1 integrates with Google Sheets for live on-chain data dashboards. Meta AI handles social sentiment correlation within Meta platforms. Kimi K2.5 has the longest context window available (1M+ tokens) for full protocol audit analysis. Grok 4 has real-time X/Twitter data access for on-chain plus sentiment fusion. The 50 prompts in this guide are copy-paste ready for immediate use.

Why this is the most important workflow in crypto trading right now

Six years ago, interpreting MVRV Z-Scores, SOPR crossovers, exchange reserve depletion, and funding rate differentials simultaneously required either a team of dedicated analysts or a quant trader with years of Python experience. The data existed. The interpretation required expertise that most traders did not have and could not afford.

In 2026, that expertise is available to every trader with a browser tab and the ability to copy and paste.

AI language models — Claude, ChatGPT, DeepSeek, Gemini, Grok, Kimi, Meta AI — have become the most powerful analytical copilots ever available to non-technical traders. Feed them the raw numbers from Glassnode. Paste the funding rate table from CoinGlass. Drop a Dune Analytics export into the chat. Ask the right questions. Receive interpretations that would previously have taken a senior analyst three hours to produce.

This guide is not theoretical. Every prompt in it has been tested. Every workflow is specific, actionable, and replicable without writing a single line of code. It covers the full spectrum from retail beginners making their first on-chain trade to professional quant traders building automated analysis pipelines.

By the end of this guide, you will have:

• A complete understanding of which AI model to use for which on-chain analysis task
• A library of 50 copy-paste-ready prompts covering every major on-chain metric
• Three tiered workflows — beginner, intermediate, and advanced — that you can implement today
• A quant/algo automation architecture for traders who want to systematise the process
• The specific on-chain data sources that feed most effectively into AI analysis

The only prerequisite is a browser. The rest is in these pages.

Part one: The AI landscape for on-chain traders in 2026

Before the workflows, understanding the strengths and weaknesses of each model matters. The wrong tool for the wrong task wastes time and produces unreliable output.

Claude (Anthropic) — Best for structured multi-signal synthesis

Claude Sonnet 4.6 is the current recommendation of Decentralised News for the core on-chain analysis workflow. Its strengths in this specific domain:

Long-context precision. Claude processes up to 200,000 tokens of context in the current version — meaning you can paste weeks of daily on-chain data, multiple metrics simultaneously, and an extended analytical framework, all in a single conversation without the model losing track of earlier context.

Structured analytical output. Claude consistently produces organised, numbered, logically sequenced analysis. When asked to identify signals, assign confidence levels, and construct trading implications, it follows the structure you specify rather than producing free-flowing prose that is harder to act on.

Calibrated uncertainty. Claude acknowledges when data is ambiguous or when multiple interpretations are equally supportable. For on-chain analysis, where false confidence is dangerous, this calibration matters. It will say “the MVRV reading is approaching the historical distribution of cycle tops but has not reached the extreme level seen in prior peaks” rather than “Bitcoin is about to top.”

No code required. Claude explains quantitative concepts, calculates derived metrics from raw data you paste, and interprets statistical relationships in plain English. It is the most accessible of the frontier models for non-technical traders.

Limitation: Claude does not have live internet access in the standard interface at claude.ai (though web search can be enabled). All on-chain data must be manually provided. This is a feature as much as a limitation — it forces you to curate the data you actually need rather than letting the model access irrelevant information.

Practical access: claude.ai — Pro subscription ($20/month) provides Claude Sonnet 4.6 with full context capacity. The free tier is functional for lower-volume analysis.

ChatGPT (OpenAI) — Best for code generation and automated pipelines

GPT-5.4 (as of May 2026) and its predecessors hold specific advantages for traders who want to build beyond manual analysis:

Python code generation. For traders who want to automate the data-to-analysis pipeline — pulling from Glassnode API, processing it with pandas, sending to an AI endpoint — ChatGPT produces reliable, runnable Python code that Claude tends to be slightly more conservative about. Ask ChatGPT to “write a Python script that pulls the last 90 days of BTC SOPR from the Glassnode API, calculates the 14-day moving average and standard deviation, and flags days where SOPR is more than 1.5 standard deviations below the mean” and you will get working code.

TradingView Pine Script. For traders who want to overlay on-chain signals on their TradingView charts, ChatGPT generates Pine Script indicator code directly. Request “a Pine Script indicator that plots a custom signal when the 7-day funding rate average is above 0.03% on Bybit — source the data manually via input — and overlays a green band on the price chart” and the code will run on TradingView.

Web browsing integration. ChatGPT with browsing can fetch current data from CoinGlass, DeFiLlama, and other platforms directly, without requiring you to copy and paste.

Limitation: GPT-5.4 occasionally produces confident-sounding analysis on on-chain metrics where the interpretation is debatable. The model tends to be less calibrated than Claude about acknowledging uncertainty. Verify important signals independently before acting.

Practical access: chatgpt.com — Plus subscription ($20/month) provides GPT-4o and access to the advanced models. API access at platform.openai.com for automation builders.

DeepSeek V4 — Best free option for quantitative reasoning

DeepSeek V4, released March 2026, runs approximately 1 trillion parameters with 32 billion active via mixture-of-experts routing. Its API costs approximately $0.28 per million input tokens — roughly 27 times cheaper than comparable closed models.

For on-chain analysis, DeepSeek’s specific strengths:

Mathematical precision. DeepSeek V4 excels at step-by-step quantitative reasoning. Feed it a table of MVRV values over 12 months and ask it to calculate the Z-score at each point, identify the periods where the Z-score exceeded 7, and compare those periods to subsequent 90-day returns. The calculations will be correct and the methodology transparent.

Cost-efficient batch processing. For traders who want to analyse hundreds of wallets, run statistical analysis on years of on-chain data, or process large CSV exports, DeepSeek’s API pricing makes it the economically rational choice for high-volume analytical tasks.

Self-hosting option. The V4 Lite variant (~200 billion parameters) can be run locally on enterprise-grade hardware, which matters for traders who are concerned about data privacy when sharing proprietary on-chain analysis with cloud-hosted AI systems.

Limitation: DeepSeek is a Chinese-hosted platform with documented concerns about data privacy and the potential for political censorship on certain topics. For standard on-chain market analysis, these concerns are unlikely to affect outputs. For traders handling sensitive portfolio information, self-hosting the open-source model eliminates the privacy concern.

Practical access: chat.deepseek.com (free), platform.deepseek.com (API).

Google Gemini 3.1 — Best for Google Sheets integration and multimodal analysis

Gemini 3.1, released in March 2026, has crossed 750 million monthly users and leads on certain reasoning benchmarks. For on-chain traders:

Google Sheets integration. If you maintain a Google Sheets dashboard pulling on-chain data via CoinAPI, CryptoQuant’s export features, or manual entry, Gemini’s native integration with Google Workspace means you can query your live data directly from the spreadsheet. Ask “summarise the last 30 days of BTC exchange inflows in my data and identify any weeks where inflows exceeded the 90th percentile” and Gemini reads your sheet without requiring data export.

Chart and image analysis. Gemini 3.1’s multimodal capabilities mean you can screenshot a Glassnode chart, attach the image to your Gemini conversation, and ask “what is the current MVRV Z-Score reading indicating and how does this compare to prior cycle tops?” The model analyses the visual chart without requiring numerical data extraction.

Limitation: Gemini responses can be verbose and slower than Claude or ChatGPT for equivalent analytical tasks. The Google Workspace lock-in means its integration advantages disappear if you work outside Google’s ecosystem.

Practical access: gemini.google.com (free with Google account), Gemini Advanced ($20/month), Google Workspace integration via paid tiers.

Grok (xAI) — Best for real-time X/Twitter sentiment fusion

Grok 4 (May 2026) has direct, real-time access to the entire X/Twitter data feed — the most valuable differentiator for on-chain traders who want to fuse blockchain data with social sentiment.

Real-time social sentiment. When you are analysing a significant on-chain event — a major whale wallet moving 10,000 BTC to an exchange — Grok can simultaneously query X to identify what key analysts are saying about the move, whether it is correlated with broader narrative shifts, and whether smart money accounts are reacting. This fusion of on-chain and social data in a single interface is unique among the major models.

Crypto community intelligence. Grok knows who the influential voices are in crypto, can distinguish between signal and noise on X, and can summarise the analytical consensus around a specific on-chain event within seconds.

Limitation: Grok’s training optimises for social media comprehension rather than rigorous quantitative analysis. Its on-chain signal interpretation is shallower than Claude or DeepSeek. Use it for the sentiment layer, not the quantitative layer.

Practical access: x.com (X Premium+ subscription required, approximately $22/month).

Meta AI (Llama) — Best for embedded workflow integration

Meta AI, powered by Llama 4, is available natively within WhatsApp, Instagram, Facebook Messenger, and the web interface at meta.ai. Its on-chain trading utility:

WhatsApp group analysis. For traders who coordinate in WhatsApp or Telegram groups, Meta AI can be added to group conversations. Paste on-chain data into the group, ask Meta AI to analyse it, and the entire group receives the interpretation simultaneously — useful for trading collectives and signal-sharing groups.

Zero-cost access. Meta AI is completely free with no subscription required for standard usage, making it the most accessible entry point for traders who want to start using AI for on-chain analysis before committing to a paid subscription.

Limitation: Meta AI has been slower to reach frontier capabilities in quantitative reasoning compared to Claude, ChatGPT, and DeepSeek. For straightforward interpretation tasks (explaining what a given SOPR reading means, summarising the implications of exchange reserve data), it is fully functional. For sophisticated multi-signal analysis and trading framework construction, the other models produce more rigorous output.

Practical access: meta.ai (free), WhatsApp, Instagram (native integration).

 

Kimi K2.5 (Moonshot AI) — Best for ultra-long context document analysis

Kimi K2.5, released by Chinese AI company Moonshot AI in early 2026, holds the largest publicly available context window of any major model — exceeding 1 million tokens. This matters specifically for:

Full protocol audit analysis. A complete smart contract audit document, a full year of daily on-chain metrics, or an entire DeFi protocol’s governance history can be loaded into a single Kimi conversation without truncation.

Research paper synthesis. Academic papers on on-chain economics, tokenomics analysis documents, and comprehensive protocol whitepapers can all be processed in full.

Limitation: Kimi’s reasoning quality on financial analysis tasks falls below Claude and GPT-5.4 in structured testing. Use it for the long-document problem specifically, and use Claude or ChatGPT for the analytical interpretation once Kimi has extracted the relevant sections.

Practical access: kimi.ai (free tier available), subscription for extended context.

Perplexity AI — Best for research-grade sourced analysis

Perplexity AI is not strictly a language model but an AI search engine that cites its sources with every claim. For on-chain analysis:

Sourced metric explanations. When you encounter an unfamiliar on-chain metric or want to understand the academic basis for an indicator, Perplexity provides cited explanations rather than the unsourced assertions that ungrounded LLMs can produce.

Current market context. Perplexity’s real-time web search means asking “what is the current MVRV Z-Score for Bitcoin and what does this historically indicate” returns a cited, current answer rather than requiring you to find and copy the data yourself.

Practical access: perplexity.ai (free with limited daily queries, Pro $20/month for unlimited).

The AI model selection matrix

Task	Best model	Second choice	Why
Multi-signal on-chain synthesis	Claude Sonnet 4.6	ChatGPT GPT-5.4	Precision, calibration, structured output
Python pipeline construction	ChatGPT GPT-5.4	DeepSeek V4	Code reliability, debugging support
TradingView Pine Script	ChatGPT GPT-5.4	Claude	Pine Script training data
Mathematical/quantitative calculation	DeepSeek V4	ChatGPT	Cost, precision, reasoning chain
Chart image analysis	Gemini 3.1	ChatGPT Vision	Multimodal capability
Google Sheets integration	Gemini 3.1	—	Native workspace integration
Social sentiment + on-chain fusion	Grok 4	Perplexity	Real-time X data access
Long document analysis	Kimi K2.5	Claude	Context window size
Real-time sourced data	Perplexity	Grok	Citation quality
Free, high-volume analysis	DeepSeek V4	Meta AI	Cost per token
Group/community analysis	Meta AI	—	Platform integration
Beginner-friendly interpretation	Claude	Meta AI	Clear, calibrated language

Part two: The on-chain data ecosystem

Before the workflows, understanding where the data comes from determines what analysis is possible.

Tier one: Essential free tools

CoinGlass (coinglass.com) is the primary free source for derivatives data — open interest, funding rates, liquidation maps, and long/short ratios across exchanges. The data updates in real time. For AI analysis, the most useful exports are the funding rate tables (copy as text) and liquidation heat maps (describe numerically in your prompt).

DeFiLlama (defillama.com) aggregates TVL, protocol fees, revenue, and chain-level data across the entire DeFi ecosystem. All data is free. Export as CSV for any protocol, chain, or time period. DeFiLlama is indispensable for DeFi-native on-chain analysis.

Etherscan (etherscan.io) is the primary Ethereum block explorer. For AI-assisted analysis, the most useful features are the whale wallet tracking (input a known whale address and export transaction history as CSV) and the token transfer events for specific smart contracts.

Dune Analytics (dune.com) — the free tier provides access to thousands of community-built dashboards covering protocols, chains, DEX volumes, NFT markets, and custom metrics. Find a relevant dashboard, export the underlying data as CSV, and paste into your AI model. The community query library is the largest free repository of on-chain intelligence available anywhere.

Tier two: Freemium platforms

Glassnode (glassnode.com) is the gold standard for Bitcoin and Ethereum macro on-chain analysis. Free account provides access to the most commonly used metrics: MVRV ratio, SOPR, exchange reserves, miner outflows, and active address trends. Studio tier ($29/month) unlocks advanced metrics including MVRV Z-Score, HODL waves, and realised price analysis. API access begins at $29/month.

CryptoQuant (cryptoquant.com) specialises in exchange-level flow data — who is sending Bitcoin to exchanges, in what quantities, from which wallet types, and when. The free tier provides basic charts. The paid tier ($49–$149/month) provides historical data exports and API access. For AI-assisted analysis, the most valuable CryptoQuant metrics are all-exchange inflows/outflows, stablecoin exchange inflows, and miner position index.

Santiment (santiment.net) combines on-chain metrics with sentiment analysis — social volume, developer activity, whale transaction counts, and token age consumed (a measure of long-dormant coins moving). The free tier provides limited data. The Growth plan ($49/month) provides full historical access and data exports useful for AI analysis.

Tier three: Institutional-grade platforms

Nansen (nansen.ai) labels over 500 million blockchain wallet addresses by type — funds, exchanges, market makers, early-stage investors, DeFi treasuries. This transforms raw on-chain data from “wallet X sent 500 ETH to wallet Y” into “a top-10 VC fund sent 500 ETH to a known OTC desk.” The Starter plan ($149/month) provides access to the Smart Money dashboard and wallet profiling tools. Exports feed directly into AI analysis frameworks.

CoinMetrics (coinmetrics.io) provides institutional-grade network data including realised cap, supply metrics, and network health indicators. Their community API (free) provides daily data. The professional tier provides real-time feeds for institutional workflows.

IntoTheBlock (intotheblock.com) uses machine learning to produce trading signals from on-chain data — including “In the Money” / “Out of the Money” holder analysis, large holder activity, and exchange flow signals. Free tier available; Pro tier ($29/month) provides full signal suite.

Part three: The beginner workflow — first AI on-chain analysis in 20 minutes

This workflow requires: a Glassnode free account, a Claude or ChatGPT account, and 20 minutes.

Step one: Obtain the data

Open Glassnode (glassnode.com) and navigate to the Bitcoin SOPR (Spent Output Profit Ratio) chart. SOPR measures whether the average Bitcoin being transacted today is moving at a profit or a loss relative to when it was last moved. A SOPR above 1 means the average coin transacted is selling at a profit. A SOPR below 1 means sellers are realising losses — historically a signal of capitulation and potential cycle bottoms.

On the chart, look at the current value and note the last 30 days of daily readings. If Glassnode’s free tier shows the chart without numerical export, read the approximate values visually and note them as a sequence: “Jan 1: 1.03, Jan 2: 1.01, Jan 3: 0.98…” For this beginner workflow, approximate readings are sufficient. The AI can work with the direction and magnitude even if the precision is imperfect.

Step two: Prepare your AI session

Open claude.ai and start a new conversation. The first message establishes the analytical context:

Beginner Prompt 1 — Context setting:

“I need you to act as a senior on-chain analyst specialising in Bitcoin market cycle analysis. You have deep expertise in interpreting SOPR, MVRV, funding rates, and exchange flow data. Your analysis should be accurate, calibrated, and clearly distinguish between high-confidence signals and ambiguous readings. Always state your confidence level.”

Claude will acknowledge the role. Now paste your data.

Step three: Paste data and ask for interpretation

Beginner Prompt 2 — SOPR analysis:

“Here is Bitcoin’s daily SOPR for the last 30 days:

[paste your data — e.g.: May 1: 1.02, May 2: 1.04, May 3: 0.99, May 4: 0.97, May 5: 0.95, May 6: 0.96, May 7: 1.01…]

Please analyse this data and provide:

1. What the current SOPR trend indicates about market participant behaviour
2. Whether the recent readings are historically associated with specific market phases (accumulation, distribution, capitulation, euphoria)
3. The significance of any SOPR crossings through the 1.0 level in this dataset
4. A confidence-rated interpretation: High, Medium, or Low confidence that this reading is actionable for a spot Bitcoin position
5. One specific trading implication — what this data suggests about risk/reward for a new position over the next 30 days”

Claude will produce a structured interpretation. The first time you run this analysis, read the explanation carefully. The goal is not just to extract a trading signal — it is to understand the reasoning so that you can evaluate whether the signal makes sense given other things you know about the market.

Step four: Cross-reference with one additional metric

A single metric is never sufficient for a trading decision. Add funding rates from CoinGlass:

Go to CoinGlass (coinglass.com), click on Funding Rate, select the major exchanges (Bybit, Binance, OKX), and note the current 8-hour funding rate for BTCUSDT. Copy the table showing the last 14 days of average funding rates.

Beginner Prompt 3 — Multi-signal fusion:

“Given the SOPR analysis you just completed, I now have the following funding rate data from the major BTC perpetuals exchanges for the last 14 days:

[paste CoinGlass funding rate data]

How does this funding rate data modify or confirm the SOPR interpretation? Are these two signals aligned or contradictory? Which should I weight more heavily in this market environment and why?”

Step five: The trading implication

Beginner Prompt 4 — Actionable output:

“Based on both the SOPR and funding rate analysis, construct a simple decision framework:

• If I currently hold no BTC position: what does this data suggest about entry timing?
• If I am already long BTC with a spot position: what does this data suggest about position management?
• What single on-chain threshold, if crossed, would most change this assessment?

Format your answer as three bullet points, one per scenario, with a maximum of two sentences each.”

The beginner workflow is now complete. You have: collected real on-chain data, framed it within an expert analytical context, cross-referenced two metrics, and extracted an actionable decision framework — all in under 20 minutes, with no code.

Part four: The intermediate workflow — multi-signal dashboard analysis

The intermediate workflow uses five on-chain metrics simultaneously and constructs a composite signal score. This workflow requires: Glassnode free/paid, CoinGlass, DeFiLlama, a Claude or ChatGPT session, and approximately 45 minutes for the first run (15 minutes on subsequent runs once the prompt template is saved).

The five-metric stack

Metric 1 — MVRV Z-Score (Glassnode): Measures how overvalued or undervalued Bitcoin is relative to its realised value, normalised by the standard deviation of the ratio historically. Z-Score above 7 has historically preceded major cycle tops. Z-Score below 0 has historically marked cycle bottoms.

Metric 2 — Exchange Net Position Change (CryptoQuant or Glassnode): Net change in Bitcoin held on all exchanges combined. Declining exchange reserves indicate accumulation (coins leaving exchanges into cold storage). Rising exchange reserves signal selling pressure.

Metric 3 — BTC Funding Rate (CoinGlass): The 8-hour rate paid by longs to shorts on perpetual futures. High positive funding (above 0.03% per 8 hours) indicates excessive long leverage and potential for a short-term correction. Negative funding indicates bearish futures positioning.

Metric 4 — Stablecoin Exchange Inflows (CryptoQuant): The volume of USDT and USDC flowing into exchanges. Rising stablecoin inflows suggest dry powder is entering the market — capital ready to buy. Declining stablecoin inflows suggest the buying pool is shrinking.

Metric 5 — Realised P&L Ratio (Glassnode or IntoTheBlock): The ratio of realised gains to realised losses across all on-chain transactions. A reading above 3 suggests the market is in euphoria — most sellers are booking large profits. A reading below 0.5 suggests panic — most sellers are taking losses.

The composite signal prompt

Intermediate Prompt 1 — Full signal stack:

“You are conducting a multi-signal Bitcoin on-chain analysis. Here are five metrics collected as of [current date]:

1. MVRV Z-Score: [value] — Current reading, 30-day trend: [up/down/flat]
2. Exchange Net Position Change (30-day): [+ or – X,XXX BTC]
3. BTC Funding Rate (Bybit, 7-day average): [X.XX%]
4. Stablecoin Exchange Inflows (7-day, USDT+USDC): [$X billion, vs 30-day average of $Y billion]
5. Realised P&L Ratio: [value]

For each metric individually: a) State what this reading indicates in isolation b) Rate the signal as Bullish/Bearish/Neutral for 30-day BTC price direction c) Assign a signal strength: Strong/Moderate/Weak

Then: d) Assess whether the five signals are aligned or contradictory e) Construct a composite signal score from -10 (maximally bearish) to +10 (maximally bullish) f) State the single most important caveat — what data would most challenge this composite reading?”

The scenario matrix prompt

Once Claude has delivered the composite signal, push it into scenario analysis:

Intermediate Prompt 2 — Scenario construction:

“Based on the composite signal score you just provided, construct three scenarios for Bitcoin price action over the next 60 days:

Scenario A (Bull case, 30% probability): What on-chain developments would need to occur, and what would the price trajectory likely be?

Scenario B (Base case, 50% probability): What is the most likely outcome given current on-chain conditions?

Scenario C (Bear case, 20% probability): What on-chain signals would suggest this scenario is unfolding, and what would the estimated price range be?

For each scenario, specify: (1) one additional on-chain metric to monitor as a confirmation signal, (2) one price level that would validate the scenario is occurring.”

The DeFi overlay prompt

For traders who also hold altcoins or DeFi positions, add the DeFiLlama TVL data:

Intermediate Prompt 3 — DeFi health check:

“I now want to add DeFi health context to the Bitcoin analysis. Here is DeFiLlama data:

Total DeFi TVL: $[X] billion (30-day change: [+/- X%]) Ethereum DeFi TVL: $[X] billion Solana DeFi TVL: $[X] billion Top protocol TVL changes this week: [paste top 5-10 protocols with their TVL and 7-day change from DeFiLlama]

Does the DeFi TVL trend confirm or contradict the Bitcoin on-chain composite signal? Are there any protocols showing unusual TVL movements that suggest either significant capital inflows (bullish for that ecosystem) or capital flight (bearish)?”

Part five: The advanced workflow — whale tracking and smart money analysis

This workflow uses Nansen (Starter or Pro plan) and is designed for traders who want to know what the most sophisticated participants are doing before price reflects it.

Identifying the wallets that matter

In Nansen, navigate to Smart Money → Token God Mode. Select BTC or ETH. The dashboard shows which labeled wallet categories are accumulating or distributing. Key categories: Smart Dumb Money (wallets that reliably call market direction), CEX Deposit (coins moving to exchanges for selling), and Top Traders (the highest-performing wallets over the prior 90 days).

Export the Smart Money flow data for the last 30 days as CSV.

Advanced Prompt 1 — Smart money flow analysis:

“I have exported 30 days of Nansen Smart Money flow data for Bitcoin. Here is the data:

[paste CSV data or summarise: e.g., ‘Smart Dumb Money wallets: net buyers of 2,340 BTC over 30 days, concentrated in the $88,000-$92,000 range. Top Traders wallets: net sellers of 890 BTC, concentrated at $98,000-$102,000. CEX Deposit wallets: 4,200 BTC moved to exchanges in the last 7 days.’]

Analyse this wallet flow data:

1. What does the divergence (if any) between Smart Dumb Money and Top Trader activity suggest about where they expect price to move?
2. The CEX Deposit activity — is the scale typical, elevated, or suppressed relative to historical patterns? What does it imply for near-term selling pressure?
3. Construct an interpretation that accounts for: (a) the wallet types most historically correlated with correct market direction, (b) whether this data is consistent with the on-chain composite signal from the previous analysis.
4. Flag any single wallet behavior in this data that you consider most anomalous and potentially most informative.”

Cross-chain whale comparison

Advanced Prompt 2 — Multi-chain whale analysis:

“I am going to give you Nansen smart money data for both Ethereum and Solana simultaneously, alongside the Bitcoin data from the previous analysis.

Ethereum smart money (7-day): [paste ETH data]

Solana smart money (7-day): [paste SOL data]

Questions:

1. Are the smart money flows across all three chains telling the same story, or is there rotation evident — capital moving from one chain to another?
2. If rotation is present, what does historical precedent suggest about the relative performance of the destination chain over the subsequent 30-60 days?
3. Which chain’s smart money data is most internally consistent and therefore most reliable as a signal? Which chain’s data contains the most noise?
4. Construct a relative performance view: given these three datasets, which asset (BTC, ETH, SOL) has the strongest on-chain case for outperformance over the next 30 days and why?”

Part six: The quant and algo trader workflow — automated AI analysis pipelines

This section is for traders who want to systematise on-chain AI analysis — removing manual data collection and replacing it with automated pipelines that surface signals without daily human intervention.

Architecture: The three-layer pipeline

Layer 1 — Data ingestion: Automated API calls to on-chain data providers. Glassnode API, CryptoQuant API, and CoinMetrics Community API all provide programmatic data access. The following Python pseudocode illustrates the structure:

python

import requests

import pandas as pd

from anthropic import Anthropic  # Claude API client

 

# Pull MVRV Z-Score from Glassnode API

def get_glassnode_metric(metric, api_key, start_ts, end_ts):

    url = f”https://api.glassnode.com/v1/metrics/market/{metric}“

    params = {

        “a”: “BTC”,

        “api_key”: api_key,

        “s”: start_ts,

        “u”: end_ts,

        “i”: “24h”

    }

    response = requests.get(url, params=params)

    return pd.DataFrame(response.json())

 

# Pull funding rate from CoinGlass API

def get_funding_rate(exchange, symbol):

    url = f”https://open-api.coinglass.com/public/v2/funding?exchange={exchange}&symbol={symbol}“

    headers = {“coinglassSecret”: “your_api_key”}

    response = requests.get(url, headers=headers)

    return response.json()

 

# Assemble data package

mvrv = get_glassnode_metric(“mvrv_z_score”, GLASSNODE_API_KEY, start, end)

funding = get_funding_rate(“Binance”, “BTCUSDT”)

Layer 2 — AI interpretation: Once data is collected, structure it as a prompt and send to the Claude or OpenAI API:

python

client = Anthropic()

 

def analyse_onchain_data(mvrv_data, funding_data, exchange_reserves):

    # Format data as structured text

    data_summary = f”””

    MVRV Z-Score (last 30 days):

    {mvrv_data.tail(30).to_string()}

    

    BTC Funding Rate (Binance, last 14 days average):

    {funding_data}

    

    Exchange Reserve Change (30-day):

    {exchange_reserves}

    “””

    

    system_prompt = “””You are a senior on-chain analyst. 

    Analyse the provided data and output a structured JSON response with:

    {

      “composite_signal”: -10 to 10,

      “primary_signal”: “BULLISH/BEARISH/NEUTRAL”,

      “confidence”: “HIGH/MEDIUM/LOW”,

      “key_observation”: “single most important finding”,

      “alert_threshold”: “the metric level that would change this assessment”,

      “30_day_outlook”: “brief narrative”

    }

    Output only valid JSON. No prose.”””

    

    message = client.messages.create(

        model=“claude-sonnet-4-20250514”,

        max_tokens=1000,

        system=system_prompt,

        messages=[

            {“role”: “user”, “content”: f”Analyse this on-chain data:\n\n{data_summary}“}

        ]

    )

    

    import json

    return json.loads(message.content[0].text)

 

# Run analysis

signal = analyse_onchain_data(mvrv, funding, reserves)

print(f”Signal: {signal[‘composite_signal’]}“)

print(f”Alert threshold: {signal[‘alert_threshold’]}“)

Layer 3 — Signal routing and action: The JSON output from the AI feeds into your existing decision logic:

python

def route_signal(signal):

    score = signal[‘composite_signal’]

    confidence = signal[‘confidence’]

    

    if score >= 7 and confidence == ‘HIGH’:

        # Strong bull signal — send to trading system

        send_alert(“STRONG_BUY”, signal)

        # Optionally: trigger position increase on exchange API

        

    elif score <= –7 and confidence == ‘HIGH’:

        # Strong bear signal

        send_alert(“STRONG_SELL”, signal)

        

    elif abs(score) >= 4 and confidence == ‘MEDIUM’:

        # Moderate signal — log for review

        log_signal_for_human_review(signal)

        

    else:

        # Weak signal or low confidence — no action

        log_weak_signal(signal)

        

def send_alert(signal_type, data):

    # Send to Telegram, Discord, or email

    # Or connect to exchange API for automated execution

    pass

No-code automation alternative

For quant traders who are not comfortable with Python but want automation, Make.com (formerly Integromat) and Zapier both support webhook-based automation:

1. Set up a Dune Analytics alert for a specific on-chain threshold (e.g., BTC exchange reserves drop below 2.3 million BTC).
2. When the alert fires, the webhook triggers a Make.com scenario.
3. Make.com formats the alert data and sends it to the Claude API or ChatGPT API.
4. The AI response is routed to your Telegram group, Discord server, or email.
5. You receive an on-chain alert with AI-generated interpretation in real time.

This no-code pipeline can be built in approximately two hours using Make.com’s visual workflow builder and costs approximately $10/month for the automation platform plus API credits.

The custom Claude Project setup

For traders who want a persistent AI analyst without building an API pipeline, Claude Projects provide a middle path:

In claude.ai, create a new Project titled “On-Chain Market Analysis.” In the Project Instructions, paste your complete analytical framework:

You are a senior Bitcoin and DeFi on-chain analyst for Decentralised News. 

 

Your analytical framework:

– Primary indicators: MVRV Z-Score, SOPR, Exchange Net Position

– Secondary indicators: Funding rate, Open Interest, Stablecoin flows

– Tertiary indicators: Active addresses, Miner outflows, Whale movements

 

Signal interpretation rules:

– Never provide definitive price predictions

– Always state confidence level

– Always identify the single most important contradicting signal

– Output in structured format: Signal, Confidence, Key metric, Invalidation threshold

 

Reference data for cycle comparison:

[Paste historical cycle turning point data]

 

Trading context:

– My primary instruments: BTC/USDT spot, ETH/USDT spot, BTC perpetuals on Bybit

– My risk framework: maximum 5% of portfolio at risk per trade

– My time horizon: 30-90 day medium-term positions

Every conversation in this Project starts with this context already loaded. When you collect new on-chain data, simply paste it and ask “update my on-chain reading with this new data.” The Project maintains continuity across sessions, allowing you to build a cumulative on-chain log without re-establishing the analytical framework every time.

Part seven: The 50 copy-paste prompts for on-chain AI analysis

These prompts are ready to use immediately. Replace [bracketed placeholders] with your actual data.

Category A: Metric interpretation (beginner)

Prompt 1 — SOPR basics:

“Bitcoin SOPR is currently [X]. Explain what this means in plain language, whether it is in the historically bullish, bearish, or neutral range, and what trading behaviour it suggests.”

Prompt 2 — MVRV interpretation:

“Bitcoin’s MVRV ratio is [X]. What does this indicate about whether current prices are above or below the aggregate cost basis of the market? Is this reading associated with cycle tops, bottoms, or mid-cycle consolidation?”

Prompt 3 — Funding rate context:

“The 7-day average BTC perpetuals funding rate on Bybit and Binance is [X%] per 8 hours. Put this in historical context: is this reading elevated, suppressed, or normal? What does it indicate about leveraged market positioning?”

Prompt 4 — Exchange reserve interpretation:

“BTC exchange reserves have changed by [+/-X,XXX BTC] over the last 30 days. Is this a significant flow relative to historical norms? What does it suggest about near-term selling or accumulation intent?”

Prompt 5 — Stablecoin dominance:

“Stablecoin market cap as a percentage of total crypto market cap is currently [X%]. How does this compare to historical periods of peak fear (high stablecoin dominance) and peak greed (low stablecoin dominance)? What does the current reading imply?”

Category B: Signal generation (intermediate)

Prompt 6 — Multi-metric bullish check:

“I want to assess whether the following on-chain conditions constitute a historically reliable bullish setup. Check each condition and rate the overall setup: MVRV Z-Score: [X], SOPR: [X], Exchange reserves 30-day change: [-X BTC], Funding rate: [X%], Stablecoin inflows: [+$X billion]. Rate each condition and provide an overall setup quality score from 1-10.”

Prompt 7 — Distribution phase detection:

“I am checking for signs of market distribution. Analyse these metrics and tell me whether they are consistent with a distribution phase (smart money selling into retail strength): Exchange inflows (7-day): [+X,XXX BTC], SOPR: [X], Large transaction count (>100 BTC): [X per day vs 30-day average of Y], Funding rate: [X%], Retail search volume: [high/medium/low].”

Prompt 8 — Capitulation signal:

“The following readings occurred in the last 72 hours. Assess whether this is a capitulation event or a staged decline: SOPR drop to: [X], Exchange inflows spike: [+X,XXX BTC], BTC price change: [-X%], Liquidation volume: [$X billion], Funding rate: now [X%] vs 7-day average of [Y%].”

Prompt 9 — Bottom formation test:

“Bitcoin has declined X% from its recent high. Test whether the following on-chain conditions match historical patterns that preceded prior cycle bottoms: MVRV: [X], Realised Price vs Spot Price: [spot is X% below/above realised], Long-term holder behaviour: [accumulating/distributing X,XXX BTC], Miner revenue trend: [up/down X% in 30 days].”

Prompt 10 — DeFi health scan:

“Analyse the following DeFi TVL snapshot for signs of systemic stress or systemic growth: Total DeFi TVL: $[X]B (30-day: [+/-X%]). Top gainers by TVL: [list]. Top losers by TVL: [list]. Stablecoin TVL in lending: $[X]B. Are there any concentration risks or unusual flows that suggest risk-off or risk-on behaviour?”

Prompt 11 — Altcoin rotation signal:

“Bitcoin dominance is at [X%] and has moved [+/-X percentage points] in 30 days. DeFi TVL is [rising/falling]. Altcoin vs BTC volume ratio: [X]. Assess whether conditions favour Bitcoin dominance continuation or an altcoin season. What on-chain metric would confirm the direction?”

Prompt 12 — Whale accumulation check:

“Large wallet addresses (>1,000 BTC) have changed their aggregate holdings by [+/- X,XXX BTC] over 30 days. Break this down: is this consistent with accumulation (buying dips), distribution (selling rallies), or rebalancing? What historical episodes does this most closely resemble?”

Prompt 13 — Exchange outflow strength:

“BTC exchange net outflows have averaged [X,XXX BTC/day] over the last 14 days. This is [X%] above/below the 90-day average of [Y BTC/day]. Assess: (1) whether this represents strong or weak accumulation pressure, (2) whether the pace is sustainable, (3) what price behaviour has historically followed similar outflow intensity.”

Prompt 14 — Miner capitulation assessment:

“Miner revenue has declined [X%] over the last 30 days. Hash rate is [up/down X%]. Miner outflows to exchanges are [+/- X% vs 30-day average]. Does this indicate miner capitulation? How does the current miner stress level compare to prior capitulation events and what happened to price in the 60 days following those events?”

Prompt 15 — Stablecoin inflow velocity:

“Stablecoin exchange inflows have accelerated over the last 7 days: Day 1: $[X]M, Day 2: $[X]M, Day 3: $[X]M, Day 4: $[X]M, Day 5: $[X]M, Day 6: $[X]M, Day 7: $[X]M. Is this acceleration pattern consistent with institutional accumulation, retail FOMO, or something else? What typically happens to BTC price in the 2-4 weeks following this inflow acceleration pattern?”

Category C: Trade construction (advanced)

Prompt 16 — Entry timing with on-chain context:

“I am considering entering a long BTC spot position. Current on-chain conditions: [paste your full data]. Construct a three-tier entry plan based on this data: Tier 1 (initial entry, highest conviction), Tier 2 (add if price declines to X), Tier 3 (maximum add if capitulation). For each tier, specify the on-chain confirmation signal that would validate the entry.”

Prompt 17 — Stop-loss calibration:

“I have a long BTC position entered at $[X]. On-chain conditions at entry: [data]. Current on-chain conditions: [data]. Assess whether the on-chain thesis remains intact or has deteriorated. Specifically: has any key metric moved in a direction that would invalidate the bullish case? What on-chain reading would confirm I should close the position?”

Prompt 18 — Position sizing with cycle phase:

“Based on the following on-chain metrics: MVRV Z-Score [X], SOPR [X], Exchange reserves [X], suggest how to size a Bitcoin long position as a percentage of a diversified portfolio. Scale your recommendation from 0% (maximum bearish) to 25% (maximum aggressive) based on the historical risk/reward implied by these readings.”

Prompt 19 — Derivatives position construction:

“I trade BTC perpetuals on Bybit. Current on-chain and derivatives data: MVRV [X], SOPR [X], Open Interest [X], Funding rate [X%], Liquidation map — heaviest clusters at [price levels]. Construct a derivatives position recommendation: (1) direction (long/short/flat), (2) leverage (1x-10x), (3) entry price, (4) take-profit, (5) stop-loss, (6) the on-chain signal that would most urgently prompt position closure.”

Prompt 20 — Portfolio allocation across assets:

“Given these on-chain readings across three assets — [BTC data], [ETH data], [SOL data] — and assuming a total crypto portfolio of $X, recommend an optimal allocation across all three. Base your recommendation purely on the on-chain evidence, not price action or narratives. Justify each allocation percentage with reference to specific metric readings.”

Category D: Risk management (all levels)

Prompt 21 — Maximum drawdown estimation:

“Based on MVRV Z-Score [X] and SOPR [X], estimate the potential downside risk for Bitcoin if these metrics return to historical median levels. Show the calculation. What percentage decline would bring MVRV back to its median? What percentage decline would bring SOPR back to 1.0?”

Prompt 22 — Black swan on-chain pre-signal:

“I want to identify the on-chain metrics that historically gave the earliest warning of major market dislocations (80%+ declines). Identify the 3-5 metrics that showed the most reliable early warning in 2018 and 2022. Apply that framework to current conditions and rate whether we are showing any early warning signs.”

Prompt 23 — DeFi liquidation cascade risk:

“Major DeFi lending protocols have the following health data: [DeFiLlama lending data — TVL, borrow rates, collateral ratios]. Identify: (1) whether any protocol shows unusual borrow/supply ratio changes that suggest growing liquidation risk, (2) the BTC or ETH price level at which cascading liquidations would likely begin, (3) whether current conditions resemble March 2020 or May 2021 DeFi stress events.”

Prompt 24 — Exchange solvency check:

“I hold funds on [exchange]. The most recent proof of reserves data shows: [data]. Analyse whether this proof of reserves is adequate, whether the liabilities appear properly collateralised, and whether there are any on-chain signals (large exchange wallet movements, outflow spikes, reserve ratio changes) that should trigger concern about counterparty risk.”

Prompt 25 — Market cycle phase identification:

“Using the following on-chain metrics as inputs — MVRV Z-Score [X], NUPL [X], Puell Multiple [X], SOPR [X], Bitcoin days destroyed [X] — identify which phase of the Bitcoin market cycle this most closely resembles: Accumulation (post-bottom), Early bull, Mid-cycle correction, Late bull (distribution), Bear market. Rate your confidence and cite the most compelling evidence.”

Category E: Protocol and DeFi analysis

Prompt 26 — Protocol TVL quality assessment:

“This DeFi protocol has the following TVL data: [DeFiLlama data — TVL, chains, top liquidity sources, token incentives]. Assess whether the TVL is ‘real’ (organic user demand) or ‘mercenary’ (driven primarily by token incentives). What percentage of TVL would likely leave if token incentives were removed? Does the TVL trend suggest growing or declining protocol utility?”

Prompt 27 — DEX volume analysis:

“This DEX has the following volume data from Dune Analytics: [data]. Is this volume consistent with genuine trading activity or does the volume/TVL ratio, trade size distribution, or wallet concentration pattern suggest wash trading or bot activity? How does this protocol’s volume quality compare to the sector benchmark?”

Prompt 28 — Yield sustainability check:

“A DeFi protocol is offering [X%] APY on [asset]. Here is the protocol’s on-chain financial data: TVL [X], protocol revenue [X/month], token price [X], token emission rate [X/month]. Assess whether this yield is sustainable from protocol revenues, from token emissions (inflationary), or from a combination. At current token price, how long before the emission-funded portion of the yield becomes unsustainable?”

Prompt 29 — Bridge flow analysis:

“This cross-chain bridge has the following flow data from Dune: inflows [X] last 7 days, outflows [Y] last 7 days, dominant asset [Z], origin chains [list], destination chains [list]. What does this bridge flow pattern suggest about capital rotation between chains? Is this consistent with risk-on (moving to more speculative chains) or risk-off (moving to safer chains) behaviour?”

Prompt 30 — Protocol competitor comparison:

“Compare these two competing DeFi protocols using their on-chain data: Protocol A: TVL [X], daily fees [Y], unique users [Z], token FDV [W] Protocol B: TVL [X], daily fees [Y], unique users [Z], token FDV [W] Which protocol has stronger fundamentals on a per-dollar-of-TVL basis? Which has the better revenue-to-FDV ratio? Which is growing faster and at what rate?”

Category F: Macro and institutional analysis

Prompt 31 — ETF flow interpretation:

“Bitcoin spot ETF net flows over the last 10 trading days: [data — daily inflows and outflows per fund]. Identify: (1) whether institutional sentiment is accumulating, distributing, or neutral, (2) whether the flow pattern is consistent with strategic allocation or tactical trading, (3) how ETF flow data has historically correlated with short-term price direction.”

Prompt 32 — On-chain vs macro divergence:

“Bitcoin on-chain data suggests [bullish/bearish] based on [metrics]. However, macro conditions (US dollar strength/weakness: [data], 10-year yields: [data], equity market risk-on/risk-off: [data]) suggest [opposing direction]. Analyse this divergence: which has historically been the more reliable leading indicator — on-chain fundamentals or macro conditions? In what time frame does each typically play out?”

Prompt 33 — Mining economics analysis:

“Current Bitcoin mining economics: Hash rate [X EH/s], Mining difficulty [X], Block reward [X BTC], Average miner cost of production (Cambridge estimate): $[X]/BTC, Current price: $[X]. Are miners currently profitable or underwater? At what price does the marginal miner break even? What does the relationship between current price and miner cost of production historically imply for price floor?”

Prompt 34 — Institutional vs retail divergence:

“On-chain data shows the following breakdown: Wallets >1,000 BTC: net purchased [X BTC] in 30 days. Wallets 0-0.1 BTC: net sold [X BTC] in 30 days. This represents: institutional accumulation while retail distributes, or retail accumulation while institutions distribute. Which pattern currently applies, and what does historical precedent suggest about subsequent price direction when smart money and retail are on opposite sides?”

Prompt 35 — Cycle timing projection:

“Using the following cycle indicators — MVRV Z-Score [X], Puell Multiple [X], Bitcoin Cycle Master Indicator [X] — and assuming the current halving occurred in [month/year], estimate: (1) how far into the current cycle are we expressed as a percentage, (2) what the historical distribution of returns is from this cycle phase to the expected peak, (3) the implied peak timing range based on prior cycle structures.”

Category G: Quant and algo applications

Prompt 36 — Signal backtest request:

“I want to backtest a simple on-chain trading rule: ‘Go long BTC when SOPR crosses above 1.0 from below; close the position when SOPR exceeds 1.04 or when MVRV Z-Score exceeds 6.’ Using historical data that I will provide, estimate: (1) the number of signals generated in the last 4 years, (2) the average win rate, (3) the average holding period, (4) the risk/reward ratio. [Paste historical data.]”

Prompt 37 — Metric correlation analysis:

“Here are 12 months of daily data for three metrics: SOPR, BTC/USD price, and Exchange Net Position. [Paste CSV data.] Calculate or estimate: (1) the Pearson correlation between each metric pair, (2) whether any metric leads the others (test 1, 3, 7-day lags), (3) which metric has the highest predictive correlation with 14-day forward returns.”

Prompt 38 — Anomaly detection:

“Here is 90 days of daily BTC funding rate data from Bybit: [data]. Identify: (1) the mean and standard deviation, (2) any days where the rate exceeded 2 standard deviations from the mean, (3) what happened to BTC price in the 7 days following each anomaly, (4) whether the current reading is within, near, or beyond the normal distribution.”

Prompt 39 — Python code for on-chain signal:

“Write Python code that: (1) calls the Glassnode API to retrieve daily SOPR and MVRV Z-Score for the last 365 days, (2) calculates a composite signal score (SOPR deviation from 1.0, normalised to -1 to +1) + (MVRV Z-Score, normalised to -1 to +1 using historical min/max), (3) outputs a daily signal table as CSV. Include error handling and comments.”

Prompt 40 — TradingView Pine Script for funding alert:

“Write a TradingView Pine Script indicator that: (1) takes a user-input value representing the current 8-hour BTC funding rate, (2) displays a green background when funding is below 0.005% (under-leveraged), yellow when between 0.005% and 0.02% (normal), red when above 0.02% (over-leveraged), (3) plots a label showing the annualised funding rate in percentage terms.”

Category H: Altcoin and specific-protocol analysis

Prompt 41 — Altcoin on-chain fundamental score:

“Using the following on-chain data for [token]: Daily active addresses [X], Transaction volume [X], Developer commit activity [X/month], Token holder distribution (top 10 holders own [X%]), Exchange inflow/outflow [X], Smart contract interactions [X/day]. Score this token’s on-chain fundamentals from 1-10 against the sector median and identify its single strongest and single weakest metric.”

Prompt 42 — New token launch analysis:

“A new token launched [X] days ago. On-chain data from Dune: Initial holder count [X], Current holder count [X], Top 5 holder concentration: [X%], Liquidity pool depth: $[X], 24-hour volume/TVL ratio: [X]. Assess: is this a healthy, organically growing token distribution, or does the data suggest: (a) team/insider concentration, (b) bot/sybil attack on distribution, (c) wash trading on volume?”

Prompt 43 — Token unlock impact assessment:

“Token [X] has an upcoming unlock of [Y%] of total supply on [date]. Current circulating supply: [X]. Current market cap: $[Y]. Largest unlock recipients: [team/investors/ecosystem]. Historical precedent: prior unlocks for this token or comparable tokens resulted in [X% average price decline]. What does on-chain holder behaviour in the 30 days before this unlock suggest about how affected parties are positioning?”

Prompt 44 — NFT market on-chain health:

“NFT market on-chain data (Dune): Total NFT sales volume (7-day): $[X]M vs 30-day average $[Y]M. Top collections by volume: [list with 7-day volume]. Unique buyers (7-day): [X] vs 30-day average [Y]. Floor price trends for blue-chip collections: [data]. Is the current NFT market showing signs of accumulation, distribution, or capitulation? What does holder behaviour in the top collections suggest?”

Prompt 45 — Layer 2 adoption metrics:

“Analyse these L2 adoption metrics from L2Beat and Dune: [chain] TVL: $[X]B (30-day change: [X%]). Daily transactions: [X] (30-day trend: [up/down X%]). Unique active addresses: [X]. Bridge inflows from Ethereum: $[X]M (7-day). Developer activity: [X GitHub commits/month]. Which L2 shows the strongest organic growth indicators? Which metric is most predictive of sustained adoption versus speculative inflow?”

Category I: Market sentiment and psychological indicators

Prompt 46 — Fear vs greed on-chain check:

“Crypto Fear & Greed Index is currently [X]. Cross-reference this with these on-chain sentiment proxies: Coin Days Destroyed [X vs 30-day average], Short-term holder SOPR [X], Exchange inflows from new wallets [X]. Does the on-chain data confirm the Fear & Greed reading, or is there a divergence? When on-chain sentiment diverges from the F&G index, which has historically been more predictive?”

Prompt 47 — Social vs on-chain divergence:

“Social media sentiment for Bitcoin is [positive/negative/mixed] based on the following indicators: [Santiment social volume, LunarCrush sentiment score]. Meanwhile, on-chain data shows: [key metrics]. When social sentiment and on-chain reality diverge, what does historical data suggest about which resolves first — sentiment catching up to on-chain, or on-chain reversing toward sentiment?”

Prompt 48 — Greed peak identification:

“Assess whether the following combination of signals represents a greed peak — a point where market euphoria has created maximum risk: MVRV Z-Score [X], Funding rate (annualised) [X%], Short-term holder profit [X%], Social volume [X vs baseline], Options skew [X], Stablecoin ratio [X%]. Rate this as 0-10 on the historic greed scale and compare to known peak conditions from previous cycles.”

Prompt 49 — Recovery phase classification:

“Bitcoin has recovered [X%] from its cycle low. On-chain conditions: Long-term holder accumulation rate [X BTC/month], Short-term holder cost basis vs spot: [spot is X% above/below short-term holder realised price], Exchange reserves trend: [declining/rising], New address growth: [X% monthly]. Classify this recovery: early accumulation, genuine bull market resumption, or dead-cat bounce. What is the single most important differentiating metric?”

Prompt 50 — The AI analyst daily briefing prompt:

“You are my daily on-chain analyst. I will provide you with a standard data package each morning and I want a concise briefing in three sections: (1) Critical signal — the most important on-chain development from the last 24 hours. (2) Trade implication — one specific, actionable observation relevant to my BTC/ETH positions. (3) Watch list — two metrics to monitor today, with the level at which each would become significant. Data for today: [paste daily on-chain summary].”

Part eight: Where to execute when the analysis is complete

On-chain AI analysis is intelligence. Execution requires an exchange. For derivatives traders acting on funding rate signals, whale accumulation reads, and MVRV cycle analysis, the following platforms integrate directly with the workflow this guide describes.

For perpetuals and leveraged positions based on on-chain signal output, Bybit provides the deepest BTC and ETH perpetuals liquidity with competitive maker fees critical when entering positions sized on the basis of on-chain conviction. BloFin offers competitive taker fees for traders using on-chain analysis to time precise entries on derivatives. For institutional-scale execution where on-chain signals are informing positions above $100,000 in notional, GRVT provides the hybrid settlement model that eliminates counterparty risk on large leveraged positions.

For spot accumulation during on-chain-identified accumulation phases (exchange reserve depletion, MVRV approaching lows, SOPR bottoming), Binance and OKX offer the deepest spot liquidity for large BTC purchases with minimal slippage. MEXC provides 0% maker fees for limit-order-disciplined spot buyers — the natural execution method for on-chain-informed accumulation strategies that target specific price levels rather than chasing market orders.

For DeFi execution following on-chain TVL and protocol flow analysis, gTrade on Arbitrum and Polygon provides decentralised perpetuals access for traders who prefer non-custodial execution. deBridge enables efficient cross-chain capital movement between the chains tracked in the multi-chain on-chain analysis workflows above.

For hardware wallet storage of spot positions accumulated during on-chain-identified cycle bottoms — the most powerful application of this entire guide — Ledger provides the self-custody layer that transforms on-chain analysis into long-term wealth accumulation rather than simply trading intelligence.

Part nine: The integrated daily routine

The full power of AI on-chain analysis comes from routine. A daily 20-minute practice of data collection → AI interpretation → signal comparison builds the pattern recognition that makes exceptional readings immediately identifiable.

Morning routine (15 minutes): Collect seven numbers: current MVRV Z-Score, current SOPR, 7-day average BTC funding rate, 24-hour exchange inflows, stablecoin inflows, DeFi TVL 24-hour change, and BTC price change. Open Claude with your saved Project context. Paste the seven numbers. Ask for the daily briefing (Prompt 50). Read the output.

Weekly deep dive (45 minutes): Every Sunday, collect the full intermediate five-metric stack (section four), export the Nansen smart money summary, and run the full multi-signal analysis. Construct the three-scenario matrix (Prompt 7). Compare this week’s composite signal to last week’s. Note whether the signal has strengthened, weakened, or reversed.

Monthly calibration (60 minutes): Once per month, review whether the AI’s signal interpretations correlated with subsequent price action. This is not about whether the signals were right — it is about identifying which metric combinations in your specific workflow are most reliable and which are producing noise. Feed this calibration data back into your Claude Project as reference material.

The final word

The traders who will dominate crypto markets over the next decade are not those with the fastest computers or the largest teams. They are those who combine on-chain intelligence with AI synthesis — who understand what the blockchain is actually saying while the majority of the market is watching price charts and reading news.

This workflow costs under $40 per month in subscriptions. It requires no coding skills. It takes 20 minutes per day. The edge it produces — knowing that exchange reserves are depleting while funding rates are neutral while smart money is accumulating — is the same edge that institutional analysts charge six figures annually to produce.

The data exists. The AI exists. The prompts are in these pages.

Start with Prompt 50 tomorrow morning.

This article is for educational purposes. AI analysis of on-chain data is an interpretive tool, not a predictive guarantee. All trading decisions involve risk. This does not constitute financial advice.

Affiliate disclosure: Decentralised News maintains affiliate relationships with Bybit, BloFin, GRVT, Binance, OKX, MEXC, gTrade, deBridge, and Ledger. Links in this article include affiliate codes. This does not influence the editorial content.

Published by Decentralised News | Author: Heath Muchena | May 2026

Recommended reading: 

7 On-Chain Metrics That Predicted Every Major Crypto Bottom (And What They Show Now)

The Algorithm Is the Market: How AI Now Controls Global Finance and Where Capital Goes Next

Tokenized US Treasuries: Best On-Chain Yield Products Compared in 2026

AI Trading Agents On-Chain: Which Protocols Are Generating Real Returns and Which Are Yield-Farming Theater

AI Agents That Trade Crypto for You in 2026: What They Can Actually Do, What They Can’t, and Where the Money Is Being Made