Market Intelligence Agent

Overview

A signal-generation agent that produces technical analysis and macro intelligence. It doesn't present output to an end user — it feeds the intelligence layer that the article-drafting pipeline draws on at the Research stage. Like the other agents in the system, integration with the pipeline was designed in from day one, not added later.

Problem

Editorial pipelines that cover markets need a reliable upstream data layer. The alternative — starting each draft with a manual research pass — creates latency, inconsistency, and a bottleneck that doesn't scale. The research stage needs structured, reusable signal input, not freeform notes.

TA and macro signals are also different in character: technical analysis operates on price structure and momentum; macro operates on economic conditions, rates, and cross-asset relationships. A single model prompt handles neither well. They need separate generation paths that produce compatible output.

Solution

Two signal-generation paths with a unified output interface:

• TA signals — price structure, key levels, momentum indicators, and pattern recognition synthesised into structured signal objects. Timeframe-aware; outputs at the resolution the downstream pipeline expects.
• Macro signals — economic conditions, rate environment, sector and cross-asset context assembled into a parallel signal layer.
• Intelligence layer — both signal paths feed a shared layer that the article-drafting MCP queries at the Research stage. The drafting agent doesn't call market data directly; it reads from the intelligence layer.
• Day-one pipeline integration — the interface between this agent and the drafting pipeline was defined before either was fully built. No retrofitting.

Tech Stack

Outcome

TA and macro signal generation operational. Intelligence layer feeding the article-drafting pipeline at the Research stage. Signal output consumed without manual intervention — research stage runs on structured input, not freeform prompts.