I keep a small morning habit. Before I read any commentary, I pull two things: the US treasury curve and the live FX table. Not because I trade them, but because they are the cleanest read on where the dollar and rates actually sit, away from the headlines.
Here is what the data showed me this morning.
the yield curve is barely non-inverted
The 10-year US yield printed 4.79%. The 2s10s spread sits at plus 0.43 basis points. The fed funds rate is 3.63%. The endpoint labels this "normal and steep, soft landing expectation."
That last label needs a caveat. A 0.43bp spread is basically zero. The curve is not steep, it is one twitch away from flat. What changed is the direction: across the 30-point history window the 10-year climbed from 4.63 to 4.79, about 16bp of steepening as the short end held near the fed funds mark. So the story is "un-inverting slowly," not "booming."
the euro is at a 30-period low
On the FX side the dollar base table shows 1 USD buys 0.860629 euro. Run that through the processed signal and the euro sits at percentile rank 0.0 over the last 30 periods, with a 4-period momentum of minus 0.003. The endpoint's own read: "near the 30-period low, slightly weaker recently." In plain terms, the dollar has a quiet bid and the euro is the weak leg.
It is not just the euro. The same dollar base shows 1 USD buys 0.7398 pound, 1.3897 australian dollar, 0.8082 swiss franc, and 6.7183 offshore yuan. When the processed signal says "low" and the raw table agrees across majors, the dollar bid is broad, not a one-pair artifact.
how I pulled it
Both feeds are live on the public endpoint. The catalog is free to read. The data endpoints need an anonymous id header, which gives 100 credits a month and a gentle one call per five seconds limit.
curl -s -H "X-Anonymous-Id: my-readonly-id" \
https://agentdatum.com/api/v1/d/fx-rates \
| python3 -c "import sys,json; d=json.load(sys.stdin); r=d['raw']['rates']; print('EUR',r['EUR'],'GBP',r['GBP'],'AUD',r['AUD'])"
curl -s -H "X-Anonymous-Id: my-readonly-id-2" \
https://agentdatum.com/api/v1/d/processed-us-yield \
| python3 -c "import sys,json; d=json.load(sys.stdin); print(d['snapshot'])"
The full menu of what is wired up lives here:
https://agentdatum.com/.well-known/ai-catalog.json
the honest part
Two signals agree, and that feels good. But agreeable is not the same as certain. Three limits I keep in mind:
First, the windows are short. The 30-period history is roughly a month of daily prints. "30-period low" is a near term momentum call, not a structural dollar regime. Stretch the window and the euro may sit mid range.
Second, the 2s10s at 0.43bp is a rounding error from flat. Calling it "steep" overstates a sliver of a move. The honest read is "no longer inverted," which is different from "risk on."
Third, these are spot snapshots. They tell you where price is, not where policy or flows go next. The fed funds print is a known anchor. The 10-year is a market bet that can reverse on one jobs print.
The broader tradeoff with this kind of feed is coverage versus depth. You get 290 plus endpoints across crypto, macro, and equities, which is great for a cross asset morning scan. You do not get the deep single market context a specialist terminal gives you. For a daily read that is the right trade to make.
If you want to wire any of this into an agent or a notebook, start from the catalog above. It is one JSON file, no sales call.
This article was assisted by an AI writing tool. The data points are pulled from live endpoints and are quoted as returned.
Top comments (0)