Is there a free API or open dataset for AI data center credit stress?

Short answer: Yes. Glitch in the Credit publishes a daily 0 to 100 AI credit stress index, its 17 signals and six layer scores, and supporting data, as JSON and CSV under CC BY 4.0. The latest reading needs no key: api/latest.json. The full dataset needs a free key, which comes with the daily email. As of Oct 27, 2026 the index reads 50 (Elevated), with 2 of 6 layers flashing.

What’s in it?

Dataset	Files	What it holds
Latest reading	JSON	Index, band, change, layers, chain reaction, signals, BDC summary, tender-offer counts, grid, recent events
signals	JSON · CSV	The signals: status, trend, scores, source, evidence date
history	JSON · CSV	Layer scores and statuses for every daily check, from Oct 25, 2026
events	JSON · CSV	Dated events behind the readings, with sources
macro	JSON · CSV	Government credit series: OFR stress index, Fed excess bond premium and GZ spread, loan officer survey, bank loans to nonbanks (credit the OFR or Federal Reserve)
filings	JSON · CSV	SEC 8-K red flags, late-filing notices, BDC tender offers and results, six months
bdcs	JSON · CSV	Listed BDCs: NAV per share, PIK share, shares outstanding, from SEC XBRL
grid	JSON · CSV	Data center hookup and capacity readings, with sources
collateral	JSON · CSV	Our GPU cost inputs (power, space, chip prices, lives, cost of capital) and the floors worked out from them, each with source and date

The endpoint list is machine-readable at api/index.json.

How do I get today’s reading?

curl https://glitchinthecredit.com/api/latest.json

The part to read first is stress_index. On Oct 27, 2026 it was:

{"value": 50, "scale": "0-100", "band": "Elevated", "change_since_previous": -2, "previous_check": "2026-10-26"}

Other top-level fields include as_of, generated_at, summary, chain_reaction, layers, signals, bdcs, tender_offers_90d, license and methodology.

How do I get a key for the full dataset?

The file api/v1/full.json needs a free key.

Subscribe at newsletter.
Open the key link at the bottom of any daily email.
Send the key as Authorization: Bearer <key> or add ?key=<key>.

Limits: 10 requests a minute and 500 a day per key. The CSV and JSON files under /data/ are open and don’t need a key.

Can an AI agent read it directly?

Yes, without scraping. The site publishes llms.txt (every signal with its status, plus data links), llms-full.txt (the full daily brief as Markdown), and Dataset structured data on the home and data pages. robots.txt allows GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Claude-User and other named AI crawlers.

What is the license?

Our index, layer scores, signal readings and archive are CC BY 4.0: credit “Glitch in the Credit” and link to the page you used. Government series keep their publishers’ credit lines. Third-party prices (stock closes, GPU rents) are shown on the site with credit and are not offered as downloads.

What other free sources exist?

SEC EDGAR XBRL APIs: filings, BDC schedules, company facts. Free; 10 requests per second; send a User-Agent with an email.
SEC BDC Data Sets: quarterly XBRL extracts, listed as an open government data asset.
OFR Financial Stress Index: daily, from the US Treasury’s Office of Financial Research, with a data download.
FRED API: thousands of series, free with a key, with an attribution line required. Some series are copyrighted by their owners.

What are the limits?

The daily history starts on Oct 25, 2026, and the overall index has values from Oct 26, 2026. There’s no long back-history yet.
Written readings are judgment calls based on public reporting, some of it secondhand, and each links its source. The stress index averages graded statuses; it isn’t a probability.
The index reports what the readings show. It isn’t a forecast, a timing tool or financial advice.

Follow the debt. Not a timing tool. Not financial advice.

Sources: Glitch in the Credit data page, api/index.json and methodology, all checked Oct 27, 2026; SEC and OFR pages linked above.