sugra.blog

Blog

Why Your AI Assistant Needs a Data Connector

An AI assistant knows how prices and markets work, but not this month's numbers. A data connector lets it look them up while it answers, with the source and the date. Two questions we asked with Sugra API attached: why coffee costs about double what it did in 2021, and what happened to TLT since 2020.

Published

The coffee question, 34 seconds. Every price on screen is written out below, with its source and its date.
Field Notes lockup. Headline: Why your AI assistant needs a data connector. Below it: Two questions, answered with data: why coffee doubled since 2021, and what hit TLT since 2020. On the right, a chat bubble reads Why is my coffee so expensive? A dashed line runs to an amber node labeled Sugra API, data connector, which branches to two cards: coffee in US stores, $9.30 a pound, BLS, August 2026, with a rising sparkline; and TLT since July 2020, -54.5%, monthly close, September 2026, with a falling sparkline.

Ask an AI assistant why coffee got so expensive and it will explain how coffee prices work. Ask it what a pound of coffee costs this month and it can only guess, because this month’s numbers were not in anything it learned from.

A data connector closes that gap. The assistant looks the numbers up while it answers, and gives you each one with the source and the date. Sugra API is a data connector for Claude and ChatGPT: 1,600+ endpoints across 36 domains, drawn from 160+ primary sources such as statistics offices, central banks and international agencies. You add it once, sign in, and ask in plain words. There is no API key to copy.

The clip above is the first of two questions we asked with Sugra API attached. The second is what happened to a Treasury bond fund after 2020.

Why is my coffee so expensive?

The question, as a coffee drinker would type it: “Why is my coffee getting so expensive? Is it going to get cheaper next year?”

With Sugra API attached, the answer starts from the price itself:

  • Ground coffee in US stores cost $9.30 a pound in August 2026, against $4.59 in January 2021: about double. That is the average price published by the US Bureau of Labor Statistics.
  • Behind it are the beans. Arabica cost $3.59 a pound on the world market in July 2026, against an average of $1.33 in 2018, in the IMF’s monthly commodity prices.
  • Cocoa went further: $10,710 a tonne in January 2025, against about $2,300 in 2018. It fell to $3,241 by March 2026, then climbed back to $5,619 in July.

Ground coffee in US stores, price per pound, every month from January 2021 to August 2026: $4.59 in January 2021, a peak of $9.72 in April 2026, $9.30 in August 2026. October 2025 has no value in the series and is left as a gap.

Data: Sugra API - one request, nothing else: GET /api/v1/fred/series/APU0000717311?observation_start=2021-01-01.

For the second half of the question, the numbers point both ways. The store price peaked at $9.72 in April 2026 and has been a little lower every month since. Arabica is below its November 2025 high of $4.10, but it rose from $3.08 in June to $3.59 in July. Whether coffee gets cheaper next year depends on harvests that have not happened yet, and an honest answer says so.

What happened to TLT after 2020?

TLT is a fund that holds long-term US Treasury bonds. We asked Claude, with Sugra API attached, what happened to it after 2020. It pulled the fund’s price history back to 2002, the Treasury yields behind it and the fund’s own list of bonds, and came back with three findings.

It lost more than half its price. TLT closed July 2020 at $171.00 and September 2026 at $77.78, down 54.5%. Counting the interest it paid out along the way, the loss since 2020 is 44.8%.

Most of the damage came from real interest rates, not inflation fears. When long-term rates rise, bonds already issued lose value, and long bonds lose the most. The 30-year Treasury yield went from 1.31% in July 2020 to 5.37% in September 2026, in monthly averages. More than three of those four points came from the real yield, what the bond pays after inflation, which went from -0.29% to 3.09%. Expected inflation, the gap between the two lines below, grew by less than one point.

30-year US Treasury yield, monthly average, January 2019 to September 2026. Nominal yield in amber, real yield after inflation in grey, the gap between them shaded. July 2020: 1.31% nominal, -0.29% real. September 2026: 5.37% nominal, 3.09% real.

Data: Sugra API - two requests, nothing else: GET /api/v1/fred/series/DGS30?observation_start=2019-01-01&frequency=m and the same for DFII30.

The hedge stopped working. From 2002 to 2020, in the 20 months when SPY, a fund that tracks the S&P 500, fell more than 5%, TLT rose 3.9% on average. Since 2021 there have been six such months, and TLT fell in all six.

This explains what happened; it is not advice on what to buy.

Add it once

In Claude, open the Sugra API listing and press Connect to Claude. In ChatGPT, open the Plugins tab and find Sugra API, or go straight to the listing. Either way you sign in to Sugra once, and there is no key to copy.

The Free plan gives 50 requests a day, and every plan reaches every endpoint. A question like the coffee one fits easily. The TLT one does not: looking inside the fund meant three lookups for each of its 46 bonds, 138 requests for that step alone.

For that there is now Personal: $9 a month for 1,000 requests a day, made for people who use Sugra API through their assistant rather than from code. You sign in the same way, and there is still no key to manage.

Every price and yield in this post was computed from data Sugra API returned.

Integrate with one key across every product direction.

Get API keyDocsFamily