SoftSignal Chat

The analyst you always
wanted. Built on real data.

Generic AI guesses at commodity market data. SoftSignal Chat reads the actual numbers before answering — live COT positioning, current USDA figures, this week's EIA report, real weather deviations.

Open Chat Interface →
The problem with generic AI

ChatGPT doesn't know what happened last Friday.

General-purpose AI assistants are trained on snapshots of the internet taken months or years ago. When you ask about corn positioning or Brazil arabica stress conditions, they answer from memory — and their memory is stale, approximate, and sometimes simply wrong.

Commodity markets move on weekly data releases: CFTC reports every Friday, EIA every Thursday, USDA monthly. A research tool with a knowledge cutoff is a research tool that can't do the job.

This is not a limitation you can work around by asking better questions. If the data isn't there, the answer isn't there.

The difference

Same question. Very different answer.

Generic AI (ChatGPT, Gemini, etc.)
COT positioning Estimates from training data. May reference positioning from months ago or cite the wrong contract.
Weather & crop stress General knowledge of growing regions. Cannot provide current CWSI deviation or soil moisture readings.
USDA supply & demand May know the structure of the WASDE report but not the current month's figures or the revision vs. prior month.
EIA storage Cannot access this week's report. Will estimate or refuse.
Cross-market analysis Can reason across markets but with no live data, the reasoning has no foundation.
SoftSignal Chat
COT positioning Reads actual CFTC disaggregated data through the most recent Friday release. Returns net position, percentile rank vs. full history, and week-over-week change.
Weather & crop stress Pulls ERA5-Land agro-climatic data for six growing regions — current CWSI deviation, soil moisture, rainfall, temperature — all vs. climatological normals.
USDA supply & demand Reads the current WASDE balance sheet with ending stocks, stocks-to-use, and the revision vs. the prior month's release.
EIA storage Reads this week's actual EIA report — storage level, injection/withdrawal, 5-year average, and percentile rank vs. history.
Cross-market analysis Can query multiple data sources in one response — compare corn COT against HRW drought and WASDE ending stocks simultaneously.
What MCP actually means

A technical standard. A practical advantage.

MCP stands for Model Context Protocol — an open standard developed by Anthropic that lets AI assistants connect directly to live data sources instead of relying on training memory.

For you, it means one thing: when you ask SoftSignal Chat a question, it reads the actual data before answering. The assistant calls the same data pipeline behind our weekly reports — live, structured, sourced from public authoritative releases — and reasons over it in real time.

No hallucinated numbers. No outdated figures. No "as of my knowledge cutoff." The answer is grounded in the data or the assistant tells you it doesn't have it.

What you can ask

Questions that actually get answered.

Where is managed money positioned in KC arabica right now, and how does it compare to the past three years?
Pulls this week's CFTC disaggregated COT, computes the net position, and returns a percentile rank vs. full history since 2010 — not an estimate.
How is Brazil arabica weather tracking versus normal for this time of year, and does the seasonal rank suggest unusual stress?
Reads actual ERA5-Land CWSI deviation for the current dekad and compares it against the same calendar period across all prior years in the dataset.
Give me a cross-market COT scan — which commodity markets are at extreme short positioning right now?
Queries all 14 covered markets simultaneously and returns sorted percentile ranks — identifying which are historically oversold or overbought in one call.
Is natural gas storage above or below average, what does ENSO say about summer demand, and where is managed money positioned?
Three separate data sources — EIA storage, NOAA ENSO status, and CFTC COT — combined into a single coherent answer.
How it works

Live data. Every session.

1

You ask a question

Plain English. No special syntax. Ask exactly what you'd ask a research analyst.

2

SoftSignal reads the data

The assistant selects the right data sources — COT, weather, WASDE, EIA, drought — and pulls current figures. Data refreshes hourly from SoftSignal's pipeline.

3

You get a grounded answer

The response is built from the actual data, not from memory. Numbers are real. If data is unavailable, the assistant says so.

What data is covered

COT positioning (CFTC disaggregated) for 13 commodity markets plus ICE EU Robusta · Agro-climatic conditions for 6 growing regions (Brazil, Colombia, Ethiopia, Vietnam, Indonesia) · USDA WASDE supply & demand for corn, soybeans, wheat, and cotton · EIA weekly natural gas storage · US Drought Monitor for commodity belts · USDA NASS planted acreage · NOAA ENSO status. Data is sourced from public authoritative releases and updated throughout the week.

Use your own AI

Already use Claude or ChatGPT? Connect SoftSignal directly.

Every paid subscription includes the SoftSignal MCP data connector — the same live commodity data, available inside the AI assistant you already use. Add it once, then ask your questions there instead of in our chat window.

Server URL — paste this when adding a custom connector:

https://chat.getsoftsignal.com/sse

Two-minute setup guides for each client:

Claude.ai (web)

Setup guide →

Claude Desktop

Setup guide →

ChatGPT

Setup guide →

Prefer not to set anything up?

Our hosted SoftSignal Chat needs no configuration — same data, in your browser, included with the Analyst tier.

Privacy & data

Your questions are your business.

SoftSignal Chat does not store conversation content. Your questions and the assistant's responses exist only in your browser during the active session. When you close the tab, the conversation is gone — we have no record of what you asked.

Your email address is used only to verify your active subscription. It is never used for marketing without your consent and is never shared with third parties outside the service providers required to operate the platform.

For full details: Terms of Use & Privacy Notice.

Ready to ask better questions?

Four weekly intelligence reports, the live MCP data connector, and — on the Analyst tier — the AI research assistant.

Subscribe →
Research — $99/mo · All weekly reports + MCP data connector
Analyst — $150/mo · Everything in Research + AI Chat portal (25 queries/day)
Already a subscriber? Open the chat interface →