Give your AI agent eyes and hands on the LoRa band.
LoRaScope is agent-readable and agent-controllable by design — live packet stream, structured capture queries, and inject/emulate over telnet and REST.
Most LoRa analysers are a screen a human stares at. LoRaScope isn’t. It exposes its live feed and its controls to an AI agent — or any script — over telnet and REST. An agent can watch the band, read decoded packets, run custom decoders, then inject and emulate traffic to test a hypothesis and watch the result. Monitor, reason, act, observe — in a loop, autonomously.
It works at the radio layer directly, so an agent can instrument a site even where there’s no LoRaWAN network server and no gateway to log into.
One analyser, every flavour of LoRa — LoRaWAN, LoRaBASIC, MeshCore, and more — all surfaced through the same single REST and telnet API. An agent switches protocol the same way it switches channel: one call, no separate tool per stack.
Every packet is logged to SD as human-readable CSV and raw frame captures, and those files are reachable over the same API — so a script or AI can pull logs, replay them, and mine historical traffic without a laptop on site.
<div class="lh-shot-cap">See</div>
See — agent-readable
Stream the live decoded-packet feed over telnet or REST: RSSI, SNR, source, payload, timestamps. Query the capture store, link health, and bandwidth/duty-cycle stats as structured data. No screen-scraping, no proprietary client — just the bytes on the air, machine-readable.
<div class="lh-shot-cap">Act</div>
Act — inject & emulate
An agent doesn't just observe — it can transmit. Inject test packets, replay captured traffic, and emulate a hub, repeater, or sensor node to reproduce a fault or stress a link. Build timed sequences and canned payloads to exercise a gateway exactly the way real traffic would.
<div class="lh-shot-cap">Loop</div>
Loop — capture & replay
Save every frame to a capture store, then replay it offline. An agent iterates on a fault — watch, guess, inject, observe — re-running the same traffic until the problem is cornered. Reproducible by construction, not "wait for it to happen again."
Custom decoders the agent can run
Your packets aren’t always standard. LoRaScope runs your own JavaScript plugins against every packet on the air — custom payload decoders, site-specific alerting, exports to the system of your choice. An agent invokes the same decoders, so it reads your packets, not just raw LoRa frames. Drop a plugin in, reload, and it’s live on the same feed as everything else.
No gateway needed
Works at the radio layer — an agent can instrument a site with no LoRaWAN network server and no gateway to log into. Just the analyser and the band.
Closed-loop testing
Monitor → reason → inject → observe, autonomously. Reproduce faults and verify fixes without a human in the loop or field hardware on the bench.
Scriptable surface
Telnet for streaming, REST for queries and control. Point any agent or automation at the URLs and go — no proprietary SDK, no vendor lock-in.
Reproducible
Capture store plus replay means a fault captured once can be replayed forever — by you, by an agent, or by a regression harness on every firmware build.
In the field — letting an AI debug a LoRa network
This isn’t a roadmap. It’s a real session.
I hit a massive run of packet corruption on a live network. I downloaded the logs from the analyser and watched fully auto-decoded packets stream in real time — RSSI, SNR, source, payload, the lot — until the pattern became visible.
Once I could see what was on the air, I pointed the AI at the analyser’s telnet and REST URLs. It monitored the live feed itself, read the capture store, and fired its own pretend transmits to test hypotheses about what was corrupting the frames. No hand-holding — it watched, guessed, injected, and observed the result in a loop.
In parallel I used Emulation: the analyser pretended to be a HUB, then a Repeater, while I tested a real node against it. I could reproduce the fault and exercise the link without touching field hardware — the corrupted traffic was captured once and replayed as many times as the investigation needed.
The net of it: the AI closed the loop on a real LoRa fault — watch, reason, inject, observe — using the analyser as its instrument. That’s the pitch. LoRaScope is the radio-layer tool an agent can actually drive.
Where it fits with the rest
- The Scope — the analyser itself: capture, decode, emulate, the custom-JS-plugin runtime the agent invokes.
- The Gateway — its secure remote-access tunnel lets an agent drive a Scope on a site you’re not standing in. Monitor a remote LoRa network from anywhere.
- The Expertise — AI-assisted debug harnesses and automated regression testing against a replayed traffic corpus, built for your site.