Questions

What does Pressure measure?
Pressure summarizes conditions across the global macro field each day, on a five-word scale: expansion, loosening, neutral, tightening, compression. One reading per UTC day. It describes conditions, nothing else.
Are these predictions?
No. They are measurements. Nothing here forecasts anything or recommends any action.
Why is the raw score not published?
The five-word reading is the product. The underlying score and the method that computes it stay inside the instrument; publishing them would let the measurement be reconstructed. What is published: each day's reading and its date, the last five days on this site, and thirty days of history on Pressure Atlas Pro.
How does an AI agent connect?
Add the MCP endpoint as a connector in Claude or ChatGPT (OAuth), or create an API key for IDEs, scripts, and agents. Every client sees all seven tools; the docs page has the exact steps.
Can I read it without MCP?
Yes. The REST v1 API serves the same data with an API key (curl, scripts, jobs), and a public RSS feed carries the last five daily readings with no account at all.
How fresh is the data?
Pressure is daily. Today's reading is partial until the UTC day closes; get_latest_pressure marks the latest closed day.
What does the free plan include?
The latest reading, 200 tool calls a day, and full visibility of the catalog. The Pressure Atlas Pro plan adds the most recent 30 days, CSV export, and 10,000 calls a day.
Why five levels?
Five levels are the published granularity, designed to be read at a glance. The underlying score stays inside the instrument.
Can I put it on a wall display?
Yes. Open the home page with ?display=1 for a chrome-free view: the fabric, today's reading, and the recent log. It refreshes itself and keeps the screen awake; the Display button in the header does the same thing with browser fullscreen.
Which clients work?
Any MCP client: Claude, ChatGPT, Cursor, and agent frameworks that speak MCP over streamable HTTP.