Agents that watch the line
FIXT writes the test plan, runs the station, and catches drift before it becomes a failure.
In partnership with
Built by engineers at
Test lines still need someone watching them.
Someone writes the scripts. Someone keeps the line running through every revision. When a unit fails, someone reads the logs by hand: board, fixture, or test?
Root-causing one failure pulls in failure analysis, reliability, ATE, and quality. Four teams, four slices of the picture, one meeting to compare notes.
That doesn't scale. Not with volume, not with how fast hardware changes.
Full test coverage in days, not weeks.
Ingest
The schematic, the datasheets, the firmware. All of it.
Generate
A full test plan, with limits pulled straight from the parts.
Run
Runs the line. Stays watching through every revision.
Triage
Every failure triaged with the eye of four engineers, not one.
The model doesn't test your board. The harness does.
In software, the harness is the loop around the model. In hardware, it's the rig around the board. We build the one that is both.
Same model, same boards. With the harness, and without.
Failures classified board / fixture / test against engineer ground truth, across bring-up runs on production boards.
Swap the model whenever a better one lands. The harness is the part that holds.
An operator, not a copilot.
Most AI for hardware is a copilot. FIXT runs the station, watches the line, and knows when something is wrong.
First-pass yield · line 2 · WX-2.3 rev C
Fixture drift caught on shift 4; recalibrated after shift 6 · last 10 shifts
Runs the station
Executes the plan, cycle after cycle.
Watches the line
Holds the baseline across every shift and revision.
Knows when it's wrong
Separates a drifting fixture from a genuinely bad board.
See it running on your line.
One station, one fixture. Schematic in, a monitored test line out.