Skip to main content

What is an Agent?

An agent is a specialist module that knows how to fetch and interpret one type of market data. Instead of relying on a single general-purpose AI, ClearView deploys the right combination of specialists for every question you ask. Each agent:
  • Connects to a specific data source (Binance, EODHD, FRED, DefiLlama)
  • Fetches real, verified data in real-time
  • Structures the data for analysis
  • Returns results with full source attribution
The AI never invents data. Every number in a ClearView response comes from a verified source, and every response tells you exactly where each data point originated.

How the Pipeline Works

When you send a question, ClearView follows a consistent 4-step process:
1

AI Planner analyzes your question

A planning model reads your question and decides which agents are needed. You can see this reasoning in the “Plan” block at the top of every response.
2

Agents fetch data in parallel

The selected agents run simultaneously, each pulling data from their respective sources. Complex questions may invoke 3-4 agents at once with no added wait time.
3

AI synthesizes a response

A synthesis model reads all the agent data and writes a coherent analysis. It can only reference data the agents actually returned — it cannot generate numbers from memory.
4

You receive the answer with sources

The response includes text analysis, interactive charts, and a source attribution section showing exactly which API endpoints provided each data point.

The 9 Agents

Agents are organized into three groups based on their role in the analysis pipeline.

Crypto Data Agents

These agents fetch raw market data from crypto-native sources.
AgentSpecialtyData Source
DerivativesFunding rates, open interest, long/short positioning, taker buy/sell volumeBinance Futures
MacroCross-asset prices (BTC, ETH, SPX, VIX, DXY), stablecoin supplyBinance + EODHD + DefiLlama
SignalsLong/short positioning ratios across multiple pairsBinance Futures
On-ChainBlockchain metrics (MVRV, NUPL, SOPR, whale activity)Currently being rebuilt on BigQuery

TradFi Data Agent

AgentSpecialtyData Source
TradFiStock prices, company fundamentals, economic/earnings calendar, forex, yield curve, FRED macro indicatorsEODHD + FRED + Finnhub

Analysis Agents

These agents interpret data using structured analytical frameworks. They compute exact numbers and produce charts.
AgentSpecialtyWhat It Does
TechnicalTechnical analysis (RSI, MACD, Bollinger, ADX, ATR)Applies a 5-layer epistemic framework where the market regime (trending vs. ranging) conditions how every indicator is interpreted
AdvisorFinancial advisory (deep dives, scanner, portfolio)Comprehensive company analysis with 8 sections, natural-language market screening, multi-position portfolio review
QuantitativeStatistical analysisReturn distributions, seasonality patterns, range compression detection, volatility regime classification — all computed with pandas/numpy
StructureMarket structure (P1/P2)Directional bias from temporal order of price extremes, multi-timeframe fractal alignment, flip risk assessment

Key Principles

You don't pick agents

The AI planner figures out which agents to use based on your question. Just ask naturally.

Multiple agents run in parallel

Complex questions invoke several agents simultaneously. “How does BTC look?” triggers Technical + Derivatives + Signals + Macro at the same time.

Every data point is attributed

The Sources section at the bottom of every response shows exactly which API endpoints provided each piece of data.

Failures are transparent

If an agent cannot fetch data (API down, invalid ticker, etc.), the response tells you explicitly. You will never receive silently fabricated data.
You can combine agents by asking multi-faceted questions. For example, “Give me the technical setup, derivatives data, and P1/P2 structure for BTC” will invoke Technical, Derivatives, and Structure agents in a single query.
The On-Chain agent is temporarily unavailable while blockchain metrics are being migrated to a new infrastructure (BigQuery). Queries about on-chain data will return a clear status message rather than stale data.