See it work
Monitors across the fleet, every alert with its timeline — and the in-app AI terminal to interrogate any of it, in plain language.
For teams. And agents.
Built for monitoring and resolving software and hardware systems. Where humans and agents share the watch.
Built for monitoring and resolving software and hardware systems. Where humans and agents share the watch.
Monitors across the fleet, every alert with its timeline — and the in-app AI terminal to interrogate any of it, in plain language.
Add Plexus by pasting one prompt.
Paste this into Claude Code — or any coding agent. It picks what to track, requests an API key you approve in one click, and your dashboard builds itself from the first data. Operating your fleet then happens in the app's built-in AI.
Set up Plexus telemetry in this project.
1. Fetch https://plexus.company/agent.md and follow it as your reference.
2. Install the SDK: pip install plexus-python (Python). For other languages, send telemetry over plain HTTP as described in the reference.
3. Read the codebase and propose 5-10 metrics/events worth tracking — latency and errors on external calls, throughput, lifecycle, key domain events. Show me the list (file:line, metric, why) and wait for my approval.
4. Request an API key: POST https://app.plexus.company/api/auth/claim with {"name":"<project-name>"}. Give me the verification_url to open in my browser, then poll GET https://app.plexus.company/api/auth/claim/<device_code> every 5s until approved.
5. Save the key as PLEXUS_API_KEY in the gitignored env file and add an empty entry to .env.example. Never hardcode or commit it.
6. Instrument the approved list with px.send()/px.event() under source id "<kebab-project-name>". Never block the hot path or throw from telemetry.
7. Run it so at least one real point ships, then give me my dashboard link: https://app.plexus.company/dashboards?quick=<source-id>.Plexus is one platform across the lifecycle of a hardware fleet: Connect a device with three lines of Python. Observe it on dashboards shaped like the thing you built. Predict failures before they ship.
Hardware-native — edge buffering when the link drops, thermal overlays on real CAD, schemas you can query — not bolted onto a time-series DB. Your IP stays yours.
We design for the engineer at 2am. If the SDK doesn't feel right, we keep cooking until it does. Dashboards come after the API, not before.
Edge buffering when the link drops. Thermal overlays on real CAD. Schemas you can query. Not a TSDB with extras bolted on.
Your IP — dashboards, anomaly models, schema, firmware wrappers — stays yours.
One platform across the lifecycle. Hardware-native — edge buffering when the link drops, schemas you can query — and alerts the moment a limit is crossed or a device goes dark.
Drop the SDK into the code you already ship. Structured metrics, edge buffering when the link drops — from a bench rig in a lab to a bus on orbit.
from plexus import Plexus
px = Plexus(source_id="drone-001")
px.send("battery", 3.7)
px.send("temp", 82.4)Thermal overlays on real CAD. Fleet-wide health at a glance. Schemas you can query — not bolted onto a time-series DB.
<plx-chart series="battery" window="60s" live />
Catch a bearing running hot with a threshold alert. Get paged when a unit goes dark. Every alert keeps its timeline — an audit trail you can replay.
POST /api/monitors
{
"source_id": "drone-001",
"metric": "bearing.temp",
"threshold": { "max": 85 }
}One platform across the lifecycle. Hardware-native — edge buffering when the link drops, schemas you can query — and alerts the moment a limit is crossed or a device goes dark.
Drop the SDK into the code you already ship. Structured metrics, edge buffering when the link drops — from a bench rig in a lab to a bus on orbit.
Thermal overlays on real CAD. Fleet-wide health at a glance. Schemas you can query — not bolted onto a time-series DB.
Catch a bearing running hot with a threshold alert. Get paged when a unit goes dark. Every alert keeps its timeline — an audit trail you can replay.
The whole platform. Your hardware. One command.
Ingest, storage, dashboards, alerts — the entire Plexus stack runs on your own machines. Free for a single team, no usage caps, and no phone-home: license checks are offline by design, because half our users can't reach the internet from their networks anyway.
curl -fsSL https://plexus.company/install.sh | bash
A Plexus engineer deploys alongside your team — telemetry flowing, dashboards live, fleet in production. Weeks, not quarters.
Drop the SDK into the firmware you already ship. Three lines of Python, telemetry flowing. No setup call. Usage-priced from the first data point — add a card and go.
A Plexus engineer deploys alongside your team — telemetry flowing, dashboards live, fleet in production. Weeks, not quarters.
Drop the SDK into the firmware you already ship. Three lines of Python, telemetry flowing. No setup call. Usage-priced from the first data point — add a card and go.
Telemetry flowing from every unit, dashboards shaped like your hardware, an alert the moment something crosses a limit or goes quiet. Onboarding is a conversation — tell us what you’re flying, driving, launching, or deploying. We read every one ourselves and write back the same day.