What Is a Crypto Market Research Agent?
A crypto market research agent is an AI-powered application that continuously monitors market data, summarizes changes, and surfaces actionable opportunities — without manual intervention. Using Mithril Builder, you can build one entirely through natural language prompts: describe the markets to watch, the signals that matter, and how you want to be alerted, and the agent is generated, connected to live data, and running in minutes.
Manual market monitoring has a documented attention problem. A 2025 study by the CFA Institute found that active traders who rely on manual monitoring miss an average of 34% of significant price moves during off-hours. Automated research agents solve this directly: they run continuously, apply consistent logic regardless of time or fatigue, and surface structured summaries instead of requiring the trader to parse raw data feeds.
What Mithril Builder Provides Out of the Box
Every app built on Mithril Builder comes pre-connected to the Mithril API, which provides real-time and historical data from 7 DEX exchanges and 2 prediction markets. For a market research agent, the relevant data sources include:
- Live and historical OHLCV for all supported instruments
- Real-time order book depth and spread
- Funding rates (current, predicted, and 30-day history)
- Open interest by exchange and symbol
- Mark price, index price, and basis
- Prediction market event prices and volume (Polymarket, Kalshi)
No API key configuration, no WebSocket management, and no data normalization is required. The Builder AI understands how to query each data type and renders it in whatever format you specify in the prompt.
Building the Agent: Feature by Feature
The most effective approach is to build the agent incrementally — one feature per prompt. The following walkthrough produces a complete market research agent across six focused prompts.
Prompt 1 — Market Overview Panel
Create a market overview panel that shows the 24-hour price change, volume,
and funding rate for BTC, ETH, SOL, ARB, and OP across Hyperliquid and Paradex.
Refresh every 60 seconds. Sort by absolute price change, largest first.Result: A sortable table appears showing all five assets across both exchanges. Columns include 24h change (%), 24h volume in USD, and current funding rate. The sort logic updates on each refresh cycle, so the most volatile asset always appears at the top.
Prompt 2 — Anomaly Detection
Add an anomaly detector that flags any asset where the 1-hour volume is more
than 3 standard deviations above the 30-day average 1-hour volume. Show flagged
assets in a separate "Unusual Activity" section above the main table.Result: The agent calculates rolling statistics from historical candle data. Assets meeting the threshold move into the Unusual Activity section with a badge showing how many standard deviations above normal the current volume is. This pattern reliably surfaces breakout conditions before price moves confirm them.
Volume-based anomaly detection is well-validated in academic literature. A 2024 paper in the Journal of Empirical Finance found that volume spikes exceeding 2.5 standard deviations above the rolling mean preceded statistically significant price moves within 4 hours in 61% of cases across major cryptocurrency pairs.
Prompt 3 — Funding Rate Divergence Monitor
Add a funding rate divergence section. Compare funding rates for each symbol
across all available exchanges. Highlight pairs where the spread between the
highest and lowest funding rate exceeds 0.03%. Show the long/short trade
that would capture the divergence.Result: A new section shows cross-exchange funding rate spreads. When a divergence exceeds the threshold, the agent displays a suggested trade: "Long on [exchange A] / Short on [exchange B] — estimated capture: 0.05% per 8h funding period." This is not a trade executor — it is a research summary. Execution still requires deliberate user action via bot.mithril.money or a direct API call.
Prompt 4 — Prediction Market Correlation Tracker
Add a section that shows Polymarket event prices for any events related to
crypto regulation, Federal Reserve decisions, or ETF approvals. Display the
current probability and 7-day change. Flag any event where probability moved
more than 10 percentage points in the last 24 hours.Result: The agent queries Polymarket event data through the Mithril API and filters for relevant categories. Events with rapid probability shifts are highlighted in yellow. This section gives the research agent a macro context layer — connecting on-chain price action to the prediction market consensus on events that drive crypto markets.
Prompt 5 — Automated Narrative Summary
Add an AI-generated market summary at the top of the page. Every 15 minutes,
generate a 3-sentence summary of the most notable market conditions based on
the data visible on the page. Include any anomalies, funding divergences,
and prediction market shifts.Result: A summary panel appears at the top with a timestamp. The AI reads the current state of all data sections and generates a concise narrative: which assets are seeing unusual activity, whether funding rates favor longs or shorts, and any macro events flagged from prediction markets. This turns the dashboard from a data display into a research briefing.
Prompt 6 — Alert System
Add a browser notification system. Send a notification when:
1. Any anomaly is newly detected (not already flagged in the previous cycle)
2. Any funding rate divergence newly exceeds 0.05%
3. Any prediction market event probability moves more than 15 points in an hour
Include the asset name and the specific trigger condition in each notification.Result: The agent adds state diffing logic — comparing the current data snapshot to the previous one to identify new events rather than re-alerting on persistent conditions. Notifications arrive in the OS notification tray with structured messages: "SOL-USD-PERP: Volume anomaly — 4.2σ above 30-day average" or "BTC funding divergence: Hyperliquid 0.06% vs Paradex 0.01%."
Extending the Agent: Advanced Research Features
The six-prompt agent above covers the core research workflow. These additional prompts extend it further:
Open Interest Tracking
Add an open interest chart for BTC and ETH showing the last 7 days of OI
across Hyperliquid and Paradex. Overlay price on the same chart. Flag periods
where OI increased while price fell (potential short buildup).Liquidation Heatmap
Add a liquidation level estimator for BTC-USD-PERP. Using the current
price and typical leverage distribution, show estimated liquidation clusters
at 5%, 10%, 15%, and 20% below and above the current price.Historical Opportunity Log
Add a log table that records every anomaly and divergence the agent has detected
in the current session, with timestamp, asset, signal type, and price at detection.
Export button to download as CSV.These extensions are each one prompt. The incremental build pattern means the agent grows in capability without any individual session becoming overwhelming. Total build time for the complete agent across all nine prompts is typically under 25 minutes.
Deploying and Sharing the Agent
Once built, click Deploy in Mithril Builder to publish the agent to a permanent URL. The deployed agent runs continuously — it does not require the Builder tab to remain open. Share the URL with teammates for a shared research dashboard, or keep it private as a personal monitoring tool.
For developers who want to automate actions based on agent signals — for example, placing a trade when a funding divergence is detected — the exported code can be combined with the Mithril API's order execution actions. See the full API reference at api.mithril.money/docs and additional automation examples on the Mithril blog.
Frequently Asked Questions
Does the research agent place trades automatically?
Not by default. The agent built in this guide is a monitoring and research tool — it surfaces opportunities but does not execute them. To add trade execution, you need to explicitly prompt Builder to add order placement calls to specific alert conditions, or use bot.mithril.money for pre-built automated trading bots.
How much historical data is available for anomaly detection?
The Mithril API provides up to 365 days of OHLCV history for most supported exchanges and symbols. For statistical baselines like rolling averages and standard deviations, 30-day windows are the most commonly used and are well within the available data range.
Can the agent monitor prediction markets alongside crypto prices?
Yes. Polymarket and Kalshi data is available through the same Mithril API endpoint as DEX price data. Prompt 4 in the walkthrough above demonstrates exactly this — combining on-chain DEX data with prediction market event probabilities in a single dashboard.
What happens if the Mithril API is temporarily unavailable?
Builder-generated apps should include error handling for failed API calls — request this explicitly: "Show a 'data unavailable' indicator if the API call fails, and retry every 30 seconds." Without explicit error handling in the prompt, the AI may generate apps that display stale data silently on API failure.
Can I run multiple research agents for different asset classes?
Yes. Each Builder session creates an independent app. You can build one agent focused on BTC/ETH majors, another on Solana ecosystem tokens, and a third on prediction market macro events. All three deploy to separate URLs and run concurrently without interference.