Features
Everything that already works, grouped by task and by hardware family.
Four hardware families on one screen. The algorithm is recognised from the model: SHA256 and Scrypt are counted separately, each in its own units.
MicroBT
Bitmain · stock firmware, Vnish, AML and Pitbit
Canaan
DG1, DG1+, DG Home
The system finds the hardware by itself and doesn't lose a machine when its IP address changes.
Scans the subnets you set — or the server's own ones if the list is empty — hourly by default and whenever you press Scan now. Up to 1,024 parallel connections; the range is edited in the interface and applies without a restart.
Open in demo(Subnet scan on schedule and on demand; opens in a new tab)
Found devices land in Awaiting approval and are polled right away, so you see what they are before adding them. Auto-approval can be switched on.
Open in demo(New hardware waits for approval; opens in a new tab)
Farm networks run on DHCP: a new address doesn't create a duplicate, and identity conflicts are logged instead of being fixed silently.
Enter the MAC, serial, chassis serial, nameplate power and algorithm by hand — a scan won't erase them; if the hardware later reports its own MAC or serial, that value takes over.
A machine moved to another site disappears from the fleet and the totals in one action.
The whole fleet live: a table, a card for every machine and a year of history.
Fleet hashrate and share of nameplate, consumption, mining and switched-off machines, silent machines and open errors, peak chip temperature — plus a live alert feed and group summaries.
Open in demo(Dashboard of the whole farm; opens in a new tab)
Fresh data streams into the browser and is applied in batches once a second, so even a large fleet scrolls smoothly.
Open in demo(Live table without reloading; opens in a new tab)
Sorting, filters and paging run on the server across the whole fleet. Search by IP, MAC, serial, model, label, worker or hostname. The selection is kept in the address, so you can share it as a link.
Open in demo(22 columns, filters and search; opens in a new tab)
Hashboards with hashrate, chips, PCB serial, temperatures, HW and CRC errors, bad chips and voltage spread; the PSU, pools, network, frequencies, network drops and the last reboot reason.
Open in demo(Device card down to the chip; opens in a new tab)
Hashrate and power for 6 hours, a day, a week or a month. Raw points are kept for 7 days, 5-minute averages for 30 days and hourly ones for a year.
Open in demo(Charts and a year of history; opens in a new tab)
Totals and a chart for the current filter, the selected rows or a group.
Consumption includes the standby draw of machines switched off by the operator, while W/TH is calculated only from the running ones.
Totals are kept per algorithm, each in its own units — TH/s or GH/s — with a switch in the header.
Fast every 45 seconds (summary and status) and deep every 5 minutes (hashboards, pools, PSU, error codes); both are configurable. Silent machines are retried with growing pauses and marked offline after 3 failed polls in a row.
14 alert types with thresholds that know the vendor, the group and the machine — and no alarms about what the operator switched off on purpose.
A rule is looked up from the machine to the nearest group up the tree, then the vendor, then the whole fleet. Three severities; the open window is set from 1 minute to 24 hours, and an alert closes once the condition has been gone for half of it.
A machine switched off by the operator is a status, not an outage. Hysteresis, deduplication, per-device mute, a higher temperature limit for WhatsMiner, drops measured against the machine's own norm, and silence on stale data.
Acknowledging doesn't close the incident: the engine closes it once the condition is gone, so nothing gets lost.
Open in demo(Acknowledge and dismiss, with the author; opens in a new tab)
“205 (Power supply current error)” instead of a bare number: 209 decoded WhatsMiner codes plus the 9000–9099 range, Antminer states and board failures, Avalon error bits and Elphapex self-tests — in every interface language.
Open in demo(Error codes in plain words; opens in a new tab)
Notifications are switched off per code and per vendor, from the list of codes your farm has actually reported. Codes carrying a board number collapse into one setting.
“Faulty” and “Faulty while running” are set automatically, even if notifications for that fault are off. You decide which alerts and codes count as a fault.
Open in demo(Faults tracked even when muted; opens in a new tab)
Windows by weekday and time, across midnight, in your time zone — for the fleet, groups with subgroups or single machines, for chosen alert types or codes. If the problem outlives the window, the alert opens right away.
A browser alarm on every workstation when many machines go silent, lose their pools, drop to zero hashrate or start getting rejects at once: from 2 to 10,000 machines within 1–60 minutes.
12 rules plus 2 alerts from auto-reboot. Defaults are what a fresh install starts with; every threshold can be changed for the fleet, a vendor, a group or a single machine.
| Alert | Severity | What it catches | Default |
|---|---|---|---|
| Rules | |||
| Not responding | Critical | The machine hasn't answered several polls in a row. | 3 polls, 180 s |
| Zero hashrate | Critical | A machine that is switched on reports no hashrate. | 300 s |
| Hashrate drop | Warning | The last 15 minutes fall below the machine's own norm over the 3 hours before. | below 80%, 900 s |
| Hashboard missing | Critical | Fewer hashboards are hashing than the machine should have. | 300 s |
| Overheating | Critical | Peak chip temperature is above the limit. | 90 °C; 100 °C for WhatsMiner |
| Fan | Critical | A fan has stopped. | 300 s |
| Power supply | Warning | The PSU input voltage is gone or its fan has stopped. | 300 s |
| Error code | Warning | The hardware reports an error code; codes caused by the shutdown itself are ignored. | 60 s |
| Pools unreachable | Warning | None of the machine's pools is alive. | 600 s |
| Pool rejects | Warning | Rejected and stale shares over 30 minutes exceed the limit, with at least 30 shares in the window. | above 3% |
| Firmware mismatch | Info | The firmware differs from the most common one for this model. | Off by default |
| Pools changed | Warning | Pools were changed bypassing the system; the alert stays for a day. An all-changes mode also reports changes made in the system, with the author. | Off by default |
| Auto-reboot | |||
| Auto-reboot impossible | Warning | A reboot was needed, but the miner credentials are missing or were rejected. | Raised by auto-reboot |
| Auto-reboot did not help | Critical | The series of reboots is used up and the cause remains; the machine is left for a human. | Raised by auto-reboot |
Telegram for the shift chat and for the person on duty, even from a closed network.
Alerts go to a shared chat and privately to people who linked their account with a one-time code, with separate severity thresholds for each. The bot token is stored encrypted, and a Test button checks the connection.
Each recipient picks a language with /lang. The bot answers only linked accounts and knows /status, /alerts and /help.
A button under the alert message: after a tap the message shows who took it, and the audit log records it as coming from Telegram.
Messages are gathered in a window from 10 seconds to 1 hour (60 seconds by default); a digest arrives every 12 hours by default and also proves the service is alive; recoveries can be sent too. Each message has the model, serial and worker, plus a link to the device card when the server address is set.
Telegram warns 14, 7, 3 and 1 days before the licence term ends.
Where the farm network can't reach Telegram directly, paste a VLESS, Trojan or WireGuard link as a whole — the server parses it. Only bot traffic goes through the proxy; polling stays in the local network. The proxy comes with the Linux build.
From one machine to hundreds — with confirmation, an abort button and an outcome for every machine.
Reboot, power off and on, power mode, power limit, target frequency, LED and pool change. The set depends on the family and firmware — see the hardware cards above; reboot, LED and pool change work on all of them.
Open in demo(Up to 8 operations per machine; opens in a new tab)
Confirm by typing the number of machines — the server checks it. 8 machines at a time, live progress with an abort button, and the hardware's own reply quoted for every machine.
Open in demo(Bulk actions on up to 500 machines; opens in a new tab)
Commands that interrupt mining ask for a separate confirmation. Single commands are open to operators, bulk ones only to administrators.
Keep a list of web-admin credentials: they are tried in turn, and the one that works is remembered for each machine. A machine can have its own pair. Passwords are encrypted and never sent back; worker passwords aren't stored at all.
Open in demo(Miner passwords handled for you; opens in a new tab)
Every single and bulk action with “N of M” progress, a per-machine breakdown and the author, auto-reboot included. The device card shows what was done to that machine.
Auto-reboot that won't finish off the hardware.
Off until the owner turns it on. Once a minute it checks open alerts and codes; triggers are zero hashrate, hashrate drop, missing hashboard, overheating, fan, PSU, unreachable pools and vendor error codes. Out of the box only zero hashrate is selected.
Open in demo(Auto-reboot by alerts and error codes; opens in a new tab)
Up to 3 reboots in a series by default and up to 10 within the safe range; more only with an explicit risk acceptance recorded in the audit log, and never above 50. 30 minutes between attempts by default (at least 10), and the series resets after 24 hours without the cause.
When the same cause shows up on many machines at once — 10 by default — it's a farm-wide problem, and auto-reboot pauses instead of rebooting the whole farm.
Machines switched off by the operator, decommissioned or not responding are skipped; unapproved ones and those in service only by separate switches; groups are excluded together with subgroups. The log is kept for half a year.
The farm as it really is: sites, containers, shelves and the repair cycle.
Site, block, container, shelf, row or other. A machine can sit in several groups; the primary one sets its thresholds.
How many slots a shelf, container or site has, and how many of them are occupied and free — nested groups included.
Groups as rectangles, machines as cells and 4 colour maps: state, share of nameplate, temperature against its limit, and alerts. Administrators arrange the layout with the mouse.
Open in demo(Farm layout with heat maps; opens in a new tab)
Faulty → Approval → Service center → Ready → back to work, or Decommissioned. Tables by repair step with buttons to move on, and each machine's pool account shows whose hashrate is idle.
Pools, accounts and the hashrate the pool actually credited.
Share of machines on each pool, how many are working, hashrate, how many fell back to the backup pool, and the reject rate. Machines without a login and workers shared by several machines are listed separately.
An account is the pool domain, login and algorithm — regions of one pool count as one account — with machines, hashrate, share of nameplate, power and efficiency.
For the Hash-Hut pool: the hashrate credited by the pool over the miner chart for the last day in 15-minute points, valid, stale and invalid shares, and the gap between what the pool counted and what the miner reported. It comes through the licence server — no pool token on your server.
Open in demo(Hash-Hut pool line on the chart; opens in a new tab)
Every machine keeps a history of its pool settings: first poll, changed bypassing the system, or changed by a named user. An alert can fire when pools are changed behind your back.
Open in demo(Pool history and a change alert; opens in a new tab)
Data out of the system, every action on record.
Pick fields in 6 groups and export everything under the current filter or only the selected rows. Numbers stay numbers, and the hashrate unit gets its own column.
Open in demo(Excel export with 41 fields; opens in a new tab)
Logins including failed ones, users, devices, groups, scans, settings and control — with the IP address and a filter by action.
Your server, your data, your rules.
The system runs on a server in the farm network, and its interface isn't published to the internet. History, settings and passwords stay with you.
Viewer, operator and administrator. Anyone can be limited to chosen groups with their subgroups, so a site team sees only its site. The last administrator can't be demoted, and nobody changes their own role.
Passwords hashed with argon2id, a lockout after 5 failed attempts in 15 minutes, and the same answer for an unknown login and a wrong password. Sessions last 12 hours with extension and end on a password or role change; CSRF protection on every change.
Miner passwords and the bot token are encrypted with AES-256-GCM. There is no default admin/admin: the first administrator is created with a one-time token.
On Linux the service runs under its own user in a systemd sandbox: ProtectSystem=strict, NoNewPrivileges, an empty capability set, MemoryDenyWriteExecute and PrivateDevices.
Linux or Windows, a single file, and no need for a permanent internet connection.
A single binary of about 20 MB for Linux (amd64 and arm64) and Windows with the database built in — no Docker. On Linux it installs with sudo ./install.sh, no internet needed; on Windows it runs from the unpacked folder.
The server checks for a new release every 6 hours and shows its notes. On Linux, an administrator installs it with a button: releases are signed, the signing key never sits on the update server, and a separate one-shot unit checks the signature and checksum again. On Windows, just replace hashhut-monitor.exe with the new version. No auto-update — you choose the moment.
Only the first activation needs the licence server. After that a signed lease covers 14 days, and 30 more days without contact are tolerated.
Issued in the cabinet, one per Hash-Hut account. It needs online activation once.
While the farm mines on the Hash-Hut pool — hash-hut.net as the first pool of the devices — with at least 5 PH/s SHA256 or 250 GH/s Scrypt in total, the licence renews itself: it is checked every hour, and each successful check moves the end date to 7 days from that check. It needs an activated key and a connection to the licence server.
PricingOpen in demo(Free on the Hash-Hut pool; opens in a new tab)
The same system for a computer with Windows 10 or 11: unpack the archive and run start.cmd — no installer, no command line.
Easy to read on a control-room screen and on a laptop at the rack.
Russian, English, Chinese and Thai — in the interface, in Telegram messages and in error-code descriptions alike.
Dark by default, light for bright rooms; the choice belongs to the workstation. The Hash-Hut brand style throughout.
Hashrate and power with an algorithm switch, W/TH, counters of working, switched-off and silent machines and a live-stream indicator on every screen — plus banners for the licence, a mass outage and updates.
Open in demo(Key numbers always in sight; opens in a new tab)
Dashboard, equipment, service, pools, accounts, groups, alerts, operations, auto-reboot, licence, users, audit log and settings.
Not built yet — but we'll add it when you need it. Tell us what your farm is missing.
Braiins OS, LuxOS, Hiveon, Goldshell, IceRiver, Volcminer, Bitaxe and more — we'll write a driver for your hardware.
New commands and work modes for your model and firmware.
Email, SMS, webhooks, Slack or Discord alongside Telegram.
Export to Prometheus and Grafana, API tokens for your own scripts and dashboards.
Today the mobile channel is Telegram; a dedicated app is possible.
Profitability and electricity cost per machine, group or site.
The pool line on the miner chart for pools other than Hash-Hut.
Miner firmware updates, fan and network settings.