How TradeThesis's 5-Agent Pipeline Works: Technical → Fundamental → News → Risk → Synthesis

Why One Prompt Isn't Enough

Ask a general AI chatbot "what do you think of this stock" and you get a single response trying to cover technicals, fundamentals, news, and risk all at once. In practice, that single response tends to blur categories together: a bullish narrative bleeds into the technical read, a risk factor gets buried under an optimistic headline summary, and there's no way to tell which part of the answer came from actual price data versus general pattern completion.

TradeThesis runs a different architecture: a stock's research report is built by five specialized agents, each handling one domain, feeding into a final synthesis. This is the same principle a research desk uses when it assigns separate analysts to technicals, fundamentals, and risk rather than asking one generalist to cover all three at once.

Agent 1: Technical

The Technical agent receives live OHLCV price data and computed indicator values for the stock, nothing else. Its job is to read price action on its own terms:

• Trend: price position relative to key moving averages (20 EMA, 50 EMA, 200 SMA)
• Momentum: RSI level and direction, MACD line and signal crossovers, histogram behavior
• Volatility: Bollinger Band width and position, ATR-based expected range
• Structure: support and resistance zones derived from recent price history
• Volume: whether recent moves are confirmed or unconfirmed by volume

Because this agent never sees the company's earnings, news flow, or narrative, its read can't be contaminated by a bullish story overriding what the chart actually shows, or vice versa.

Agent 2: Fundamental

The Fundamental agent works from the company's financial data: revenue and earnings trends, margin trajectory, balance sheet health, valuation multiples relative to the company's own history and its sector. It answers a different question than the Technical agent entirely: not "where is price now," but "is the underlying business improving, stable, or deteriorating."

This separation matters because a stock can look technically strong while the fundamentals are quietly weakening (a classic late-cycle setup), or technically weak while fundamentals are actually improving (a potential value setup once the price catches up). Blending the two into one read would obscure exactly this kind of divergence, which is often the most useful thing to know.

Agent 3: News

The News agent processes recent headlines and material events: earnings surprises, guidance changes, management changes, regulatory actions, competitive developments. It's built to answer:

• What happened recently that's relevant to this stock?
• Is the tone of coverage positive, negative, or mixed?
• Does the news narrative match what price has already done, or is there a gap between the two?

News sentiment that contradicts price action is a specific, useful signal that a single blended analysis would likely miss entirely, since it would just average the two into a mushy middle-of-the-road take instead of flagging the contradiction.

Agent 4: Risk

The Risk agent doesn't try to answer whether the stock will go up. It characterizes what could go wrong and how large that risk is:

• Volatility risk: ATR-based expected trading range
• Earnings risk: proximity to the next earnings date and the stock's historical post-earnings volatility
• Concentration and liquidity risk: average volume relative to typical position sizes
• Sector and macro risk: exposure to factors outside the company's direct control

Keeping risk assessment separate from the bullish or bearish case prevents a common failure mode: a strong technical or fundamental case quietly suppressing mention of the risks that would make a reasonable trader size the position smaller, or skip it.

Agent 5: Synthesis

The Synthesis agent receives all four reports and produces the final research note. Its job is not to average the inputs or restate them as a bullet list. It's built to:

• Identify where the four agents agree (a convergent signal, generally the highest-confidence setups) and where they conflict
• Weight the conflicting signals based on what typically matters more in the current context (for example, a fresh material news event usually overrides a stale technical pattern)
• State clearly what would need to change for the outlook to flip
• Present the analysis without issuing a directive buy or sell call

The value of synthesis is in surfacing agreement and disagreement clearly, not in resolving every disagreement into false certainty.

Why This Structure Beats a Single Prompt

Property	Single-prompt chatbot	5-agent pipeline
Live price and indicator data	Often absent or stale	Computed from live data
Domain separation (technical vs. fundamental vs. news)	Blended together	Enforced by design
Detects divergence (e.g., bullish news, weak chart)	Usually missed	Explicitly surfaced
Risk assessment isolated from the bullish case	No	Yes
Verifiable against the actual chart and filings	Hard to check	Each agent's claim is traceable to its source

What This Pipeline Doesn't Do

It doesn't predict where the stock is going next. No agent in this pipeline makes a price forecast, because that isn't a claim the underlying data supports (see Can AI Predict Stock Prices? for why). What it does produce is a structured, current, cross-checked view of where the technicals, fundamentals, news, and risk stand right now, and where they agree or conflict, which is the input a trader needs to build their own thesis faster than reading four separate sources by hand would allow.

Summary

TradeThesis's 5-agent pipeline exists because a single AI response covering technicals, fundamentals, news, and risk at once tends to blur exactly the distinctions that matter. Splitting the work into Technical, Fundamental, News, and Risk agents, then reconciling their outputs in a dedicated Synthesis step, produces research that's faster to generate than a manual process and more structured and verifiable than a single freeform AI answer.

---

Related reading:

• How AI Analyzes Crypto: What a 5-Agent Research Pipeline Actually Does — the same architecture applied to crypto assets
• How LLMs Are Changing Stock Research — the broader shift this pipeline is part of
• Technical vs Fundamental Analysis — the two disciplines the pipeline's first two agents are built around