RMIntelligence

Research / XAU/USD AI research copilot

Capability

An AI research copilot built for XAU/USD

Most AI chat tools answer gold questions from memory, so the price they quote can be days or months old. The RM Intelligence copilot checks the live market first. Before it answers, it reads the current XAU/USD price, computes the technicals it needs and consults our macro model, and it cites the source and time of every number.

Last reviewed 25 September 2026 · Research and decision support only, not investment advice

What it does during a conversation

The copilot runs on Claude through the Anthropic API with tool use. Claude decides which tools a question needs, our server runs them against live data, and Claude writes the answer from the results. It has three read-only tools:

Tool 1

Live gold price

The current XAU/USD reference price with its observation time, freshness state (current, delayed or stale) and source. Primary feed: gold-api.com. Backup: RM's MT5 broker terminals.

Tool 2

Gold technicals

SMA20, SMA50, RSI14, ATR14 and the 20-bar range on 1-hour, 4-hour or daily candles, computed when it is asked rather than taken from a stored snapshot.

Tool 3

Gold macro model

Daily gold returns regressed on the 10-year real yield, the US dollar index and gold implied volatility, with a rich/cheap z-score and a model fair value.

Weekly CFTC positioning, US Treasury real yields and US CPI are attached as a timestamped workstation snapshot. The copilot has to say when any of these inputs is delayed or stale.

Gold macroeconomic analysis, grounded in data

Gold is usually explained through a few macro drivers: real interest rates, the US dollar, volatility, and positioning in futures markets. The copilot checks each one against current data before it draws a conclusion:

The same inputs drive our free daily XAU/USD macro context report, so you can see what the copilot sees.

How it answers

  1. It opens with one assessment: WAIT, CONDITIONAL LONG, CONDITIONAL SHORT or INSUFFICIENT DATA.
  2. It says when the data it used was observed and flags anything delayed, stale or a proxy.
  3. It lists the evidence, the conditions, what would invalidate the view, event risk and any missing inputs, each in its own section.
  4. It tags every factual or numerical claim with a source reference.
  5. When evidence is mixed or a data feed has failed, it says so and answers WAIT. It never fills a gap with a guessed number.

Example question

Is gold rich or cheap versus real yields and the dollar right now,
and does the daily trend agree?

To answer this, the copilot calls the macro model for the rich/cheap z-score, the live price tool for the current reference price, and the daily technicals for the 20- and 50-day averages. It then checks whether all three agree. If they do not, the answer is WAIT, with the reason.

RM Intelligence Gold Research Copilot answering whether gold is rich or cheap versus real yields and the dollar
The copilot answering this question on 25 September 2026: live reference price, model build time and technicals stated first, every figure source-tagged, and a WAIT verdict because the macro model and the daily trend disagreed.

What it will not do

The copilot is available to RM Intelligence members inside the Gold Decision Lab. The live-price tool pattern it uses is documented for developers and open source on GitHub.

Further reading

See membership options