Cliff
Docs Blog Dashboard
Industrial monitoring Agents for sensor data. Anomaly detection across every machine, read and investigated by LLM agents.
01 Deploy monitors with a sentence. Describe what should worry you. Cliff drafts the monitor, marks what it inferred, and backtests it against your history before anything goes live.
Signal
analyzer-03
Signal
analyzer-07
Logic
rail_5v
Logic
dark_current
Logic
rail_sagging
Logic
optics_fouling
Alert
email
Aug 28, 02 PMAug 30, 02 PM
block_temp_clid_temp_cpeltier_current_aflow_ul_minstepper_current_a
analyzer-07 · 5 fields · 48h · normalized per series
Which analyzers are trending badly, and why?
analyzer-07 is the one to watch. Block and lid temperature have climbed together over the last two days while peltier current sits near its ceiling — the cooler is working harder to hold setpoint, which usually precedes an optics-fouling drift. Flow and stepper current are steady, so it isn't fluidics.
called 3 tools
02 Talk to your data. Ask which machines are doing badly, or why yield dropped. An agent reads the signals, works the question, and shows the trail it followed.
03 Integrate in an afternoon. Point Cliff at whatever you already have. It reads the payload, infers the shape, and starts detecting on the history that is already there.
HTTP Any endpoint that returns JSON Polled on a cadence you set. Fields are mapped from a live probe.
HISTORIAN OSIsoft, Ignition, InfluxDB Backfill years of history, then stream forward from the same connection.
BROKER MQTT and OPC UA Subscribe to topics directly. Cadence and shape are learned from the stream.
Get in touch. Tell us what you are running and we will connect one machine with you.
Cliff 2026
Docs Blog Careers Contact