The network for the Earth.

The gap.

Getting data out of the field is expensive.

Existing telemetry systems are costly. Hardware runs from $1,000 to over $5,000 per unit, plus ongoing data costs that can be as high as $600 per year. DIY solutions do exist, but are often more challenging to stand up and still require traditional cellular or satellite data plans. Many of the people closest to the land — watershed groups, small farms, tribes, researchers, teachers — are unable to use the existing option because of cost and technical roadblocks.

Mycelium is designed to make it easier, cheaper, and more reliable to telemeter sensors in the field. It uses low-cost, open-source hardware that you can build easily from off-the-shelf parts and a fully open and free software stack. While it currently suffers from some of the same issues as other DIY telemetry solutions (namely needing some technical knowledge to understand and set up the hardware), the next release will include a no-terminal setup app to streamline the entire process. The goal is to make it simple enough to allow anyone to stand up a node in a few minutes with no technical knowledge.

How it works.

Sense → Mesh → Your data.

Rather than relying on satellites or cell towers, mesh networks pass readings from one sensor node to the next, hop by hop, until they reach a gateway with internet access. Adding a node extends and strengthens the network for everyone.

Free and open-source

Designed for everyone, not just research labs.

Mycelium is designed as a public good. Parts lists, firmware, software, and assembly guides are all open source, and you build a node from off-the-shelf parts. The design is free, the radio band is license-free, and your data lands somewhere you already own.

Price it yourself — this list renders straight from the repo.

Reach.

Built for the places cell towers never reached.

The valleys, forests, and rangelands that need monitoring most are exactly where coverage ends — where the only commercial answer has been a satellite subscription. Mycelium's answer is the mesh itself: every node relays by default, so each sensor anyone deploys extends the range of the mesh for everyone. Want to monitor water quality in a deep canyon without cell coverage? Each node you deploy extends the network farther up the canyon. No WiFi or other gateways nearby? I am working on a field gateway with a cellular or satellite connection. Instead of needing a data plan for each individual sensor, one data plan for the gateway connects the entire network to the internet, even in the most remote of locations.

Your data stays yours.

The relays can't read what they carry.

Every reading is encrypted with AES-256-GCM on the sensor node and decrypted only at its owner's destination. Every other node in the mesh sees only the final address and ciphertext. This allows nodes deployed by disparate teams and individuals to reinforce each other without sharing anything but radio waves.

b6 b2 cf 9f e9 8c c2 a8 50 48 e0 59 4d 19 6c f1
27 6f ec 8c c5 18 c0 dc 89 a9 c2 30 79 0d 6a f2
87 71 28 30 cf 1f df 35 b4 d6 ff e4 22 8d 94 31
5e 52 93 45 a6 37 3c cc f2 76 13 ad 5f ba 88 2d
f7 f4 94 32 70 be b9 8c 94 23 07 cf 11 29 fb e0
50 63 a3 36 11 ab fa

2026-06-11 18:23 · leaf01 · temp_c · 18.16

Built like infrastructure, not a gadget.

Reliability and survivability are at the core of the Mycelium architecture.

Breakdowns in the field are expensive. They often necessitate a technician physically going out to the station to reset it, which is especially problematic if your nodes are in hard-to-reach locations (on mountaintops, buoys, etc.). While hardware problems are to some extent inevitable, starting with field reliability as the basis for each decision in the software stack can help make field deployments more reliable and thus cheaper.

Mycelium approaches software reliability in a variety of ways:

1. Every reading is committed to disk the moment it's taken and deleted only after its destination cryptographically confirms receipt.
2. Destination or gateway outages buffer and drain automatically. If a node loses its connection, it stores and consolidates readings until the connection is re-established. By transmitting multiple readings together after an outage, Mycelium helps to avoid network overload. If the destination is down, gateways themselves can hold onto packets, preventing retransmission from the source node and the resulting network bandwidth drain.
3. Nodes watch themselves — and recover from wedges, crashes, and even a radio knocked off the bus, on their own. Three watchdogs watch the system at separate layers (process → OS process-supervisor → silicon) and reboot if a process stalls, crashes, or enters an infinite loop.
4. Automatic notification of an offline node. If the destination script does not receive a reading for a specific amount of time (set to 3x the log rate but easily modifiable), it sends a message to the provided email that the node has fallen silent.
5. While the current node hardware design has a single logger, microcomputer, battery, and radio unit, the V2 node that is currently in the planning phase has complete hardware redundancy. The idea is that even with the physical degradation or destruction of a hardware component, the nodes should be able to reroute any process around the degraded hardware and alert the destination address of the failure. This means hardware degradation turns into something you can deal with during routine maintenance rather than necessitating an emergency trip to the field.

Features 1–4 have been verified by sabotaging the bench network with scripted failure drills:

Zero readings lost.

A real drill run against the real (bench) network, June 2026. The drills ship in the repo — run them yourself.

Where this is going.

A bench-proven core, growing toward the field.

The encrypted pipeline, the mesh, the gateway software which translates LoRa to TCP/IP, and the reliability layer all run end to end on the bench today. What comes next is making it hardened enough to survive real weather and easy enough that you never touch a terminal. Two pieces are in active development right now with two more in planning:

1. bench network running end to end ✓
2. field hardware testing ◐ in progress
3. zero-terminal setup app ◐ in progress
4. fleet dashboards ○ in planning
5. fully redundant hardware (v2) ○ in planning

All of it public, all of it free.