Turn the lab into a harness for AI
A modern lab is full of instruments, each with its own software, cables, and spreadsheet. Wendy puts a small computer beside each one and serves it to AI agents as tools, so experiments can run and report back without anyone pasting data by hand.
Concept visualizationEvery instrument becomes a tool AI can use
Put a Jetson Orin Nano, Raspberry Pi 5, or Qualcomm IQ9 next to the microscope, mass spectrometer, incubator, balance, or centrifuge.
Package an instrument’s driver as a containerized app in Python, Rust, Swift, C++, or TypeScript, and deploy it with one command: wendy run.
The app turns the instrument into an MCP server, a set of tools AI can call to capture an image, set focus, or pull a spectrum.
Each app gets only the access it declares in wendy.json, such as USB, camera, GPIO, or networking.
Concept visualizationOne secure network, with live context for AI
Wendy Net connects every device, robot, and laptop over Ethernet, Wi-Fi, Bluetooth, or Wi-Fi Aware, directly or through the devices in between.
AI agents like Claude read each instrument through MCP and can send each one its next step.
Robots carry samples between stations, on the same network and platform as the instruments.
Design, run, measure, learn, day and night
Claude designs an experiment from live readings across the lab.
Robots and instruments run it.
Every result streams straight back to the agent.
It picks the next experiment, and the next one, around the clock.
Concept visualizationRepeatable, traceable, and on site
Apps and on-device models run without internet, and you can deploy and update over USB-C or the local network.
Once a device is provisioned, every connection uses mutual TLS with per-device certificates.
Containerized apps on a versioned OS image pin each run: same code, same stack, same device state.
Apps can log inputs, model versions, and decisions to an OpenTelemetry collector on the device, for your own traceability and audit process.
WendyOS, the CLI, and the agent are Apache 2.0 open source, and every release ships with an SBOM and build provenance.
A 1:28 concept film of how a lab could run on Wendy, from the first instrument to an experiment loop that runs day and night.
3.23 s
for a Qualcomm Dragonwing IQ9 to describe two camera feeds, on device, from one timed run of the unedited demo.
Six devices on one table, completely air-gapped. The IQ9 runs a vision language model, two ESP32-S3 cameras stream to it, two Arduino UNO Qs light up when someone walks in, and an iPhone asks it what it sees.
Eight pages for lab and R&D leaders: what the disconnected lab costs, what changes with Wendy, how it holds up in a quality and security review, and a low-risk first step.
Get the brief
Common questions
Direct answers about connecting lab instruments to AI, keeping data on site, and repeatable runs with WendyOS.
A small computer such as a Jetson Orin Nano, Raspberry Pi 5, or Qualcomm IQ9 sits beside each instrument and runs its driver as a containerized app. The app serves the instrument as an MCP server, and the Wendy MCP server passes those tools to AI agents such as Claude.
Instruments your software can talk to from a device beside them, over interfaces such as USB or the local network. Your team or a coding agent writes the driver as an app and deploys it with wendy run.
Yes. Apps and on-device models run on the hardware beside your instruments, and Wendy devices can run fully offline, so instrument data does not have to leave the lab. You decide what leaves, including anything you send to a cloud AI model.
Containerized apps on a versioned OS image keep code, stack, and device state the same from run to run. Apps can log inputs, model versions, and outputs to the OpenTelemetry collector on each device, and teams can use those records in their own traceability and audit processes.
Tell us about your instruments and your workflows, and we’ll show you how Wendy fits in your lab.