A new substrate for agents
Live data, structured as signals agents can reason about.
Signal Fabric is the connective layer between operational reality and your models. It turns sensors, feeds, and APIs into typed signal objects carrying observation, context, and provenance — so agents can act on what's happening, not what happened.
What it is
A fabric, not a feed.
A feed gives you rows. A fabric gives you signals: each one carries the observation, the context it was taken in, and the provenance of the measurement. Agents can stitch signals across domains because the shape is consistent.
Weather, water, energy, space, finance, and operations all enter the fabric through the same primitives. The Dynamic Firehose API lets agents discover what's available; the Predictive Ops API lets them act on it.
The signal object
One shape across every domain.
Every signal is observation + context + provenance. Models stop guessing what a row means.
observation
what was measured
"reading": 1.42
context
where, when, what kind
"domain": "weather"
provenance
source, method, confidence
"source": "NOAA"
AI-ready by construction
Built for the agent loop.
Discovery, resolution, and pull are first-class endpoints. An agent can browse firehoses, resolve a thesis into a signal plan, pull only what it needs, and reason from typed observations.
- DiscoverGET /api/firehoses/custom — list every available stream with semantic context.
- ResolvePOST /api/thesis/resolve — translate a claim into a concrete signal plan.
- PullSubscribe only to the firehoses the plan names. No scraping.
- ActPredictive Ops endpoints expose buffers, alerts, and forecast windows.
Try it
Stage a thesis. Generate the agent prompt.
Describe a claim or question. We resolve it against Signal Fabric, show readiness, and package a prompt your coding agent can run.