The Fed, FRED, and MCP servers

Note: This is a public session with Bruce Boston and Tim Ellis. Counts below are what they read from a Federal Reserve speech or calculated on the spot. The shipping product is turbolapper - AFM (Apple Intelligence on-device; data stays on your Mac). See requirements and Technology.

They opened FRED because they already ingest it

The October 5, 2026 session starts on FRED, Federal Reserve Economic Data. Bruce and Tim have been ingesting those series with their tools, so a speech about the same corpus is on topic for them.

They point viewers to the Federal Reserve site: news, events, speeches, Waller. The speech they read is about navigating trust and AI storytelling in a world of data. On the session they treat it as that day's conference remarks, read from a script. They also say the page they opened is dated October 1.

Summary

• Corpus: FRED hosts more than 850,000 series from governments, and reached 18 million unique users last year.
• Access: an MCP server or connector, which they describe as bring your own AI. Tim says he confirmed it with a local setup, and that any system that can attach an MCP server can use it.
• Growth, as read: traffic growing 150% a year, most of it from AI and bots, the other half from people.
• Their arithmetic, labeled as theirs: that growth on 18 million is about 45 million. About 22 million humans, and about 22.5 million AI acting for humans.
• Next: Watch the session · Get the app

What the speech says FRED is

The lines they read first: FRED currently hosts more than 850,000 series from governments. It reached 18 million unique users last year. They count themselves among those users. They download the data and pull it apart.

The new piece is an MCP server, or connector, so AI systems can work directly with those datasets. They call the idea bring your own AI. Tim says he has confirmed it works locally, including with Ollama, and that it also works with any other system that can hook up an MCP server.

Bots, humans, and the 150% line

The growth line they find on the page: traffic growing 150% a year, with most of the growth from AI and bots, and the other half from flesh and blood. They read both halves as they are written.

Then they do the math out loud, and it should stay labeled as their math. If last year was 18 million, they take 150% growth to about 45 million. Split the way they split it: about 22 million humans a year retrieving the data, and about 22.5 million as AI on behalf of humans.

A second cut from the same arithmetic: even if all 18 million last year were human, humans are about 22 million now. Bruce's guess is tighter on the low side. He says human visitors are closer to 15 million then and 22 million now, which he calls about a 50% increase in human interest in the data.

The headline they pull: AI is increasing human interest in economic topics. Government data that felt closed off now feels reachable, because assistants can pull a series apart without the person first learning the whole method.

Meant for humans, read by machines

A later paragraph, as they read it: FRED was meant for humans, and AI systems interpret the data very differently. Tim's note is that the governor does not say "incorrectly."

The speech does name failure modes. AI may not cite sources of data accurately. It may also make a correlation-versus-causation error, the kind economists and statisticians already reject papers for.

That is the FRED story they stay on: a public corpus, an MCP door, a surge that is half machines and half people, and a central bank governor describing both.

FAQ

What is the MCP piece in this session?

FRED has an MCP server or connector so AI systems can work directly with the datasets. The session calls it bring your own AI. Tim says he confirmed a local hookup, and that any system that can attach an MCP server can use it.

Where do the 45 million and 22 million figures come from?

They are the hosts' arithmetic, not a second published total. The speech lines they read are more than 850,000 series, 18 million unique users last year, and traffic growing 150% a year. They scale 18 million by that growth to about 45 million, then split it as about 22 million humans and about 22.5 million AI on behalf of humans.

Does the speech say AI reads FRED wrong?

They read him as saying AI systems interpret the data very differently, not incorrectly. He does say AI may fail to cite sources accurately, and may confuse correlation and causation.

Watch and go further

• Watch: A Fed governor just talked about MCP servers
• Related: One week of analyst work · Four DGX Sparks · FRED and MCP · Community
• Product: Get the app · Technology · Pricing