Day-to-day use of Rem0te: finding a computer, connecting to it, helping someone who is not enrolled, and keeping the remote-control client current.
Screenshots are captured from a seeded demo instance. Every business, person and computer in them is fictional.
Terminology. A Business is a customer organisation and is the security boundary — a Business User only ever sees computers belonging to their business. A Computer (also called an endpoint or enrolled client) is a machine with the Rem0te agent permanently installed. See access-control.md for who can do what.
- Your email address.
- Your password.
- Sign in.
If your account has two-factor authentication enabled you are asked for a code next. Recovery codes work here too, and each one is single-use.
Below 768 px wide (a phone, or a tablet held upright) the menu is not drawn beside the page. The ☰ menu button at the left of the top bar opens it as a panel over the page; it holds everything the sidebar holds, including the business switcher, the theme and Sign out. Choosing a page closes it, and so do Esc, tapping outside it and the ✕ in its corner. While it is open, the keyboard stays inside it and a screen reader treats it as a dialog named Main menu.
Wide tables — the device list, reports, API keys — scroll sideways inside their own box rather than widening the page; swipe the table, not the page. The device workspace's tab row scrolls the same way.
Signing in lands you on the Dashboard. It answers, from stored data, what is broken, which businesses it affects, which devices need attention, what is already being worked on, and what to do next. It covers only what you can see: a Business User sees the devices granted to them, a technician sees their assignment's business and computers, a Platform Admin sees every business. A section you have no permission for is not shown at all — it is never shown as a zero.
- Site outage — several devices at one site stopped reporting together; Open alerts goes to them.
- The tiles — each is a count with what it is counted from underneath; each one links to where you act on it. Devices splits the fleet three ways: online, agent silent and offline (here 23 · 1 · 10).
- Work on next — the ranked list, every item with the reason it is there.
- Customers affected — the businesses with an outage, an alert, an offline device or a failed action.
| Part | What it is built from |
|---|---|
| Site outage strip | An open site-outage alert: several devices at one site stopped reporting together |
| Devices | Active devices (archived and not-yet-enrolled ones excluded): online, offline, and agent silent — reachable through RustDesk but the agent has not checked in for 30 minutes, so nothing it reports is current |
| Critical alerts | Open critical alerts. Alerts in a maintenance window are counted apart, not as critical |
| Failed actions | Approved-action and patch targets that failed, were refused or did not run in the last 7 days, plus any whose outcome is unknown (those block new requests until someone records what they found) |
| Awaiting approval | Plans waiting for the approval roster. Approval happens on the signer, never here; expired requests are not counted |
| Restart pending | Devices whose latest update scan says the system (Windows, or a Linux package manager) is waiting for a restart, or that have a restart pending alert. A scan older than 36 hours is flagged stale |
| Patch compliance | Of the devices with an update scan under 36 hours old, the share that list no pending updates; with no fresh scan at all it says No data. The Patch status card breaks it down (up to date, updates pending, stale scan, never scanned) with coverage and scan ages |
| Active sessions | Sessions connected now, and ones still launching. A session row that has not moved for 12 hours is not shown as live |
| Work on next | A ranked list, each item with the reason it is there (below) |
| Customers affected | Businesses with an outage, an open alert, an offline device or a failed action |
| Devices needing attention | Devices with critical or warning alerts, failed actions, offline, agent silent or restart pending, most serious first |
| Being worked on | Live sessions, acknowledged alerts (and who acknowledged them), actions in flight, patch deployments in progress |
| Recent alerts (0.55) | The newest open and acknowledged alerts: severity, computer, business, how long, and the automatic PSA ticket's reference. Each links to the alert. Needs View alerts, and lists only computers you can see. With no monitor configured it says so and links to Settings → Monitoring (Add recommended) |
| Your recent devices | The devices of your own most recent sessions |
| Recent activity | Alert, action, patch, session, device and support-request events from the audit log, last 7 days. Needs View business audit log |
How "Work on next" is ordered. A fixed ladder, then oldest first: site outages, critical alerts, actions with an unknown outcome, failed actions, warning alerts, failed client rollouts (Platform Admins), approval requests, restarts pending from a fresh scan, then info alerts. Acknowledged alerts are not in it (someone has them), nor are alerts in a maintenance window or offline alerts grouped under a site outage. Nothing is ranked by anything the dashboard cannot see, such as a customer's contract.
Every item and figure links to where you act on it — the device's Monitoring, Patches or Actions tab, the patch deployment, the alerts list. The page polls every 30 seconds and says how old its data is. Server health (CPU, memory, disk, relay) is not on the dashboard: Platform Admins find it under Settings → System status (Server health at the top right). From 0.35 the dashboard fits the window on a desktop screen (1366 × 768 and up): the figures across the top, then Work on next, Devices needing attention and Customers affected, then the map with Being worked on, Patch status, Recent alerts, Recent activity and Your recent devices beside it. A long list scrolls inside its own card instead of lengthening the page — also while a red containment strip is showing at the top (0.36: the dashboard takes the strip's height off its own). On a phone the cards stack in that order. Hide device locations (top right) folds the map away — Work on next then takes its column — and this browser remembers the choice.
- Client locations — devices placed from the address they last checked in from, clustered by city with a count. Green while an agent there is checking in, amber when the only computers there are reachable through RustDesk with their agent silent (the marker south of Portland), grey when they are all offline. The map is drawn from a bundled outline; no tile server is asked where your customers are.
- Hide device locations — folds the map away; this browser remembers the choice.
Devices lists every machine you can see: in the businesses you can reach, and — for a Business User — only the machines assigned to you or shared company-wide; for a technician, only the machines your assignment names.
- Devices — the menu entry; its tabs hold My devices, Add device and the others. From 0.64 Add device is also its own link just under Devices in the menu, for anyone with Add computers.
- Search — matches name, hostname, IP address, OS, alias, RustDesk ID, business, site and the signed-in user. The server does the matching, so it searches every page, not just the one on screen.
- The table — one row per device. Online devices come first, then by name; click any column heading to sort by it instead (click again to reverse).
- Add device — enrol a new machine (§5). Until 0.32 there were two buttons here, Enroll Client and Generate Enrollment Link; both became this one.
| Column | What it shows |
|---|---|
| Device | Name, hostname, and a dot: green online (the agent is checking in), amber reachable through RustDesk with the agent silent, grey offline. |
| Business / Site | Who the machine belongs to and where. |
| OS | An icon and the reported Windows edition or platform. |
| Health | The worst open monitoring alert, in words (0.68): Issues (a critical alert), Warning (a warning), Healthy (none, or informational ones only, and the agent is reporting) or Unknown (no warning or critical alert, but the agent is silent or the computer offline: a silent agent raises nothing). Healthy is not a promise nothing is wrong, only that no monitor says so. Click the label to open that computer's Monitoring tab (0.70). Clear, beside Warning or Issues, resolves every open alert on that computer after you confirm. It does not change any monitor, so anything still wrong is raised again, as a new alert, at the computer's next check; each cleared alert is recorded in the audit log. |
| Agent | Signed (the 0.27+ signed agent), Standard, or No agent report, with when it last checked in. |
| Last seen | Online, Reachable (RustDesk) · agent silent … in amber, or how long ago it was last heard from. |
| Logged-in user | Who the last heartbeat said was signed in. |
| CPU / RAM / disk | The last resource sample and its age. |
| Pending patches | From the last update scan: how many are waiting, and a restart tag if Windows wants one. |
| Alerts | How many monitoring alerts are open (or acknowledged). Click it for the device's Monitoring tab. |
Narrower windows show fewer columns (0.47.1, 0.48.1). So the list fits the page beside the sidebar without scrolling sideways, each of these columns appears only when the window is wide enough for it: Last seen from 1440 pixels, Agent from 1536, OS from 1700, Site from 1860 and Logged-in user from 2000. At 1280 to 1439 pixels that leaves Device, Business, Health, CPU / RAM / disk and Pending patches; a 1920-pixel screen shows everything but Logged-in user. The dot beside each name still says online, agent silent or offline. The column picker shows each one's width, for example (1700 px and wider). A column you chose is never removed, only hidden while the window is narrower than that. Alerts is off by default, because Health already shows the worst open alert (Healthy when there is none); turn it on under Columns. A browser that saved its own column choice keeps it. Device names never wrap; long business and site names are cut short, with the full name on hover; and Connect stays pinned on the right.
Every number says how old it is, and an old one is greyed. A resource sample is stale after 15 minutes (five missed heartbeats); an update scan after 36 hours. A greyed value is what the machine last said, not what it is doing now.
Health and Alerts need the monitoring:alerts permission. Without it those columns are not offered and you cannot filter or sort by them — the list says so at the bottom.
- Filters — business (when you can see more than one), site, device group, online (Online (agent checking in), Reachable, agent silent or Offline), open or critical alerts, Restart pending, agent type, and archived devices. More filters (0.34) adds platform (Windows, Linux, macOS, other), OS name contains, tag (exact), update state (pending, none pending, scan stale, never scanned), a RustDesk version range (dotted numbers, compared numerically — 1.10 is above 1.9; a computer with no reported version or a non-numeric one such as a nightly never matches a range) and last seen (not seen for N days, which includes never; seen within N days). They combine, and Clear resets them. The address bar carries the current view, so you can bookmark it; whoever opens the link still sees only their own devices. On a phone (0.35) the filters sit behind a Filters (n) button beside the search, n being how many are set; it opens a sheet with the same controls, each applying at once, and Show N devices closes it.
- Columns — choose which columns show (a Groups column is available; it
lists every group the device is in, and the CSV's Groups column does the same,
separated by
;). Compact / Comfortable sets the row height. Both are remembered in this browser. - Views (0.33) — save the current filters, sort, columns and density under a name. A view is stored on your account, so it follows you to another computer. It holds only the list's own parameters, never the devices. It is either personal (only you see it) or shared with everyone in one business who can see devices. Sharing — and changing or deleting a shared view — is for a Business Owner, a technician with computers:edit over the whole business, or the MSP, and is recorded in the audit log. A shared view's filters cannot name another business, or another business's site or group. Opening a view shows only the devices you can see: a Business User opening the owner's "everything" view gets their own devices, and a view filtering on alerts is refused to someone without monitoring:alerts. Views saved in a browser by 0.31 or 0.32 are offered for import once (Import N views saved in this browser).
- Device groups (0.33, Views → Device groups…) — named, static groups of one business's computers, for filtering and for acting on together. From 0.34 a computer can be in several groups of its business: adding it to one never takes it out of another, removing it from one leaves the rest, and deleting a group only removes that membership. The group filter shows every device in that group, and each group's count in the menu counts a device once per group it is in. An enrolment link can name a group: the new device joins it when it installs (re-enrolling an existing device adds that group and keeps its others). A device moved to another business leaves all of the old business's groups. Creating, renaming and deleting groups, and adding or removing members, is for a Business Owner, a technician with computers:edit over the whole business, or the MSP — never a Business User, whatever they hold, because a Business User may see only some of the business's computers. A group cannot hold another business's computer. Groups grant no access to anything. Every change is audited. Groups made before 0.33 that span businesses (or have no members) belong to no business: they still filter, and cannot be edited here.
- Rule groups (0.34, Views → Save filters as a rule group…) — a group whose members are whichever of its business's computers match a rule right now. The rule is the list's own filters: set them, then save. It can use site, online/offline, alerts, Restart pending, agent type and everything under More filters; search, business, group and the archived filter are never part of a rule (the group already belongs to one business, and the list's own Archived filter still applies). Nothing is copied or stored as members: the server works the membership out every time the group is read, so a computer that goes offline has left an "Online" group the next time anyone looks, with no sweep to wait for. Everyone sees only the members they can already see — a Business User sees their own computers in it — and a rule on alerts is not offered to, and refused for, someone without monitoring:alerts. A rule group is in the group filter (marked rule), the Groups column and the CSV like any other group. It cannot be added to or removed from by hand, named by an enrolment link, or turned into a static group (or back). To change it, set the filters and use Device groups… → Use current filters. Who may create, edit and delete one is the same as for a static group, and each change is audited with the rule before and after. A business can have at most 50 rule groups. If a stored rule can no longer be read, it matches nothing (never the whole business) and the groups dialog says so.
- Select and act on many (0.33) — tick rows (the header box ticks the page; the selection survives paging and filtering) and a bar offers: Add / Remove from a group, Refresh inventory (also re-checks Windows Update), Queue client update (Platform Admin only, as on the Updates page) and Export selection. At most 200 devices per request. The server checks every device on its own, with the same checks as doing it to that one device, and the result lists every device it refused and why — not visible to you, no permission, archived, Linux, no device secret, under an updates containment pause, in another business than the group, already current. A bulk refresh counts toward your limit on distinct computers per window (the fleet-abuse guard); the list stops at that limit and says so, rather than tripping it and pausing you. There is deliberately no bulk connect, archive, delete or approved action. Each bulk request is recorded in the audit log, once per business it touched, alongside each device's own entry. Filtered by a rule group (0.34), the same bar acts on the group itself: the server works out its members when you click — as you see them, at most 200 (more is refused, never cut short) — and then checks each one exactly as above. Copy into puts the computers that match now into a static group; later matches are not added. The audit entry records the group and how many it resolved to.
- Export CSV — downloads every device in the current filtered view (not just
this page), up to 10,000; narrow the filters if there are more. It needs
computers:view, contains exactly what you can see, and each export is
recorded in the audit log with its filters and row count. Export selection
is the same export limited to the ticked devices.

- Search — the device list's own search, across every page.
- Group — one of the filters; the others are business, site, online, alerts, agent, Restart pending and archived.
- Views — saved filters, sort, columns and density, personal or shared (below).
- Columns — which columns show; Comfortable / Compact beside it sets the row height.
- The row's box — ticks the device; the header box ticks the page.
- The bulk bar — appears while anything is ticked: group membership, Refresh inventory, Queue client update, Update agent, Upgrade to full agent…, Export selection.
- Views — shows the active view's name once one is open.
- A shared view — with the business it belongs to. Opening it still shows only the devices you can see.
- Save current view… — the filters, sort, columns and density, under a name; Save filters as a rule group… and Device groups… below it make and manage groups.
- Connect on a row works exactly as it does on the device page. ⋯ opens the device, or its Patches, Monitoring or Test connection tab, or copies its hostname.
There is no RustDesk ID column any more. The ID is still on the device's page, and search still finds a device by it.
Online means the agent checked in within the last 30 minutes. The agent heartbeats every 3 minutes, so a machine that has genuinely dropped shows Offline a few minutes later, not instantly. A machine RustDesk can still reach whose agent has gone quiet is neither (0.35): it is amber, Reachable (RustDesk) · agent silent with how long, and its tooltip says remote control should work but monitoring, patching and actions are not reporting — try Update agent on its Agent card. The dashboard counts these apart from online ones.
The Search box at the top of every page, or Ctrl+K (⌘K on a Mac) from anywhere, searches devices, businesses, sites, people and open alerts in one go. Type at least two characters (at most 100); results appear when you pause. ↑ / ↓ move through them, Enter opens the highlighted one, Esc closes the box. Ctrl/⌘-click opens a result in a new tab.
- The search box — at least two characters; the matched part is highlighted in each result.
- Devices — at most eight, online first among equals, with a link to the device list's own search for the rest.
- Businesses — then sites, people and open alerts, each group only if you have its list page's permission.
| Group | Matches on | Opens | Needs |
|---|---|---|---|
| Devices | Name, hostname, RustDesk ID, serial number, signed-in user, IP address, a network adapter's IPv4, alias | The device's page | View computers |
| Businesses | Name, code, email, city | The business's page | Nothing beyond being in it (as the Businesses list) |
| Sites | Name, city, address | The device list filtered to that site | View computers |
| People | First name, last name, email — "jane doe" finds Jane Doe | The Users list, filtered to that person | View business users |
| Open alerts | Title (open and acknowledged only) | The device's Monitoring tab, or the alerts list | See and work alerts |
What a device result shows. Its name, online or offline, business and site, and — unless the name itself matched — which field matched and its value (Serial: 5CG1234XYZ, RustDesk ID: 123456789, Adapter IP: 10.0.4.17). Only the field that matched is returned; the others are not sent. Every one of them is already on the device's own page for anyone who can open it, which is the only kind of device search can find.
Order and limits. An exact match comes first, then one that starts with what you typed, then one that contains it; among equals, online devices first, then by name. Each group shows at most 8. When there are more, the group says first 8; for devices there is a link to the device list's own search, which pages through everything (it matches a slightly different set of fields: OS, business and site names, but not serial numbers or adapter addresses). Archived devices and inactive sites are not found, as on their lists.
You only find what you could already open. Each group is built from the same rule as its own list page — not a copy of it: devices from the device list's scope (a Business User sees only devices assigned to them or shared company-wide, a technician only their assignment's), businesses, sites and people from their list routes, alerts from the alerts list. A group you have no permission for is left out, never shown as empty. Another business's device, site or person is not found by anything you type — not even its exact hostname or RustDesk ID. Search grants nothing: opening a result applies that page's own check again.
Privacy and limits. Searches are not written to the audit log (a keystroke
is not an event), and the API does not log the query text — an error is
logged with the path only. A reverse proxy in front of Rem0te keeps its own
access log, which records the address of each request, search included.
Results are sent with Cache-Control: no-store. Searching is limited to 120
requests a minute per client address; past that the box says Too many
searches until the minute is up. Typing pauses of 200 ms trigger a request,
so ordinary typing stays well inside it.
How fast. Measured end to end through the API on a shared scratch Postgres 16 (Platform Admin, every device in scope): 50–70 ms over 2,000 devices and 300–450 ms over 20,000 — in line with the device list's own search (56 ms and 320 ms on the same data), which scans the same way. A two-letter query that matches nearly every device (pc) is the slow case: about 0.3 s at 2,000 and 1 s at 20,000, because every match has to be ranked. There is no search index; the measured trade-off is in the 0.33 changelog.
Opening a computer gives you one page for it.
The header says who it is and whether it is reachable: name, status (Online, Offline since …, or, in amber, Reachable (RustDesk) · agent silent 12 d when RustDesk can reach it but its agent has stopped checking in), business, site and last seen, marked stale when the agent has missed its check-ins. Under it, the attention strip lists what needs looking at, each one a link to the tab that explains it: open alerts with the worst severity (→ Monitoring), Restart pending and pending updates with the age of the scan (→ Patches), a refused heartbeat or an out-of-date agent (→ Overview), and which agent the computer runs — Standard agent, Signed agent or RustDesk only (see What each agent can do).
Naming it (0.68). Rename, next to the name, sets the name the computer goes by everywhere in Rem0te and a short description shown under it. The agent never overwrites the name; the hostname is kept, shown beside it when they differ, still searchable, and Use hostname puts it back. Aliases on the Overview's Assignment card are other names it is known by (searchable), added and removed in place. The Overview's Notes card shows the latest three notes and takes a new one; all of them, and editing, are on History. All of it needs Rename/edit computers, and renames and alias changes are in the audit log.
The actions: Connect, Tools (below), Diagnose (opens Diagnostics), Run action (opens Actions; disabled, with the reason printed underneath, when the computer lacks the signed agent or no approval signer is configured), Patch (opens Patches) and More: view-only connect (technicians), the diagnostics report, Reinstall agent and Archive computer, each behind a confirmation that says what it does.
Tools (0.75). Tools, beside Connect, opens one panel listing what can be done to this computer, in eight groups: Diagnostics, PowerShell / Scripts, Services & Processes, Software, Windows Updates, Networking, Security and Maintenance. Every tool is a way into a flow that already exists — picking one opens its tab with the right thing chosen (the diagnostic's row marked, the action picked, the Security log selected) or, for scripts, the script library. Nothing runs from the panel itself: the tab's own button and confirmation do that, and the server checks your access, containment and the approval signer again exactly as before.
| Group | Tools, and the flow each opens |
|---|---|
| Diagnostics | Test connection (Diagnostics tab), Refresh inventory, Disk space, Service health, Top processes, Recent system errors (Diagnostics checks), Download diagnostics report |
| PowerShell / Scripts | Run an approved script (Automation → Scripts; windows.script.run) |
| Services & Processes | Services and Processes (their tabs), Start/stop/restart a service (windows.service.control, from a row), Restart an allowlisted service and Start the Print Spooler (Actions), End a process (windows.process.stop, from a row) |
| Software | Installed software, Install software (windows.software.install), Remove a program (windows.software.remove) — the Software tab |
| Windows Updates | Check for updates now and Restart pending? (Patches), Restart signals (a Diagnostics check), Install updates (windows.update.install, Patches) |
| Networking | Network adapters (Network tab), Network connectivity and DNS resolution (Diagnostics checks), Flush DNS / renew DHCP — coming later |
| Security | Security event log (Event logs, with Security and Information chosen), Antivirus, firewall, BitLocker, Secure Boot — coming later |
| Maintenance | Schedule a restart, Cancel a scheduled restart, Collect diagnostic logs (Actions) |
Each tool says whether it can work here, and why not:
| State | Means |
|---|---|
| Available | Opens and can be used now (in direct mode an action is signed when you request it, after a fresh sign-in) |
| Needs approval | An approved action: two approvers authorize it on the approval page before the computer runs it |
| Needs a newer agent | The computer's agent does not report the action, is older than the release that added it, or is the standard agent and the action needs the full (signed) agent |
| Unsupported here | Linux, macOS (read-only agents) or RustDesk only: nothing on the computer can run it |
| Paused | A containment pause on approved actions covers this business |
| Not reporting | The computer is offline or its agent is silent: the request waits for its next check-in (a diagnostic expires after fifteen minutes) |
| Not permitted | Your role does not allow it — it is shown without a button, so nothing you can click would be refused |
| Not set up | No approval signer is configured on this server, so no action can be approved |
| Unavailable | Archived, not enrolled, in the unassigned pool, or never bound to a device secret |
| Coming later | No flow does it yet; it is listed so the gap is visible, and nothing is shown in its place — unknown is never healthy |
Restart pending? says Yes, No or Unknown from the last update scan;
it says Unknown until a scan has answered. The panel decides only what it shows
(apps/web/lib/device-tools.ts, unit-tested by
apps/web/scripts/unit-device-tools.mjs); it adds no route and no action. Its
links can be shared: ?tab=diagnostics&check=dns_resolution,
?tab=actions&action=windows.logs.collect, ?tab=events&log=Security.
- Status — Online, Offline since …, or Reachable (RustDesk) · agent silent … when the agent has gone quiet.
- Actions — with the reason underneath when one is unavailable (here: no approval signer on this server).
- Attention strip — open alerts, pending updates with the scan's age, and which agent it runs; each a link to its tab.
- The tabs — below.
- Patch status — the Overview's summary of the Patches tab: what the evidence is and how old.
The tabs, each linkable with ?tab=:
| Tab | What is on it |
|---|---|
| Overview | Assignment (from 0.35 with every group the computer is in — static groups and the rule groups it matches right now, as you would see them in the list; each links to the list filtered by it), key alerts, patch status, the Agent card, recent activity, then the hardware, utilisation, storage, network and update cards, each with its age |
| Monitoring | Alerts first (acknowledge and resolve here), then what is watched on this computer in words, the agent's last report, and what changed |
| Software | Installed software with a filter, and recent software changes (Windows agents only); Uninstall on a program asks the full agent to remove it as an approved action (0.41 — software-deployment.md) |
| Processes | Every running process, as a snapshot with its age and a Refresh (0.33); from 0.45 End process on a row asks the full agent to end it as an approved action (approved-actions.md) |
| Services | Every service with state, start mode and run-as account; snapshot with age and Refresh (0.33); from 0.45 Start, Stop, Restart and Startup type on a row, as approved actions on the full agent, with protected services marked Protected |
| Network | Every network connection: addresses, gateway, DNS, DHCP, MAC; snapshot with age and Refresh (0.33) |
| Patches | Patch status, updates offered, Windows Update history |
| Event logs | A page of one Windows event log, on request (Windows agents only); from 0.48 Ask AI and Create ticket on an entry (ai-and-tickets.md) |
| Diagnostics | Test connection: server, computer and your own RustDesk |
| Actions | Request an approved action and see past requests; from 0.47.1 the Print Spooler's last known state, with Refresh, and Nothing to do when it is already running |
| Chat | Chat with the person at the computer, receipts and earlier transcripts; emoji, files and pictures from 0.44 (signed Windows agent; elsewhere it says why not — chat.md) |
| History | Remote sessions, technician notes, the AI-written summary (labelled as such) and who can connect |
The Agent card (Overview, 0.34) says which agent the computer runs (Standard, Signed, or the read-only Linux/macOS agent), its version and whether that is the current one, the update channel and the portal's update mode, and — for a signed agent — the result of its last update check. Its buttons:
- Update agent (standard Windows agent): re-runs the current installer at the next check-in. It is the same as Reinstall agent: the same permission, refused while updates are paused for the business, and audited.
- Check for update now (signed agent): the agent runs its own signed update at the next check-in. It installs only what the signed repository lists.
- On Linux and macOS the card prints the exact command to re-run on the computer; those agents act on nothing the portal sends.
- Upgrade to full agent (standard Windows agent): asks for a reason and a fresh sign-in, then moves this computer to the full (signed) agent at its next check-in, keeping its enrolment and RustDesk ID. From 0.37 it is also at the top of the computer's page, beside Connect, with Update agent, and the Agent out of date chip opens it. It is always a button: when it can't be used it is disabled and says exactly why — no signed release published, no device secret, the portal in Signed only mode, updates paused, a rollout running. A request can be cancelled until it completes. Devices → Agent migrations (a tab for anyone who may request one, from 0.35; the page itself needs only View computers and is linked from every Agent card) lists pending, held, not confirmed and completed requests. To move several computers, select them on All devices and use Upgrade to full agent… (at most 200; each one is refused with its own reason). See Update signing → Moving existing agents.
Processes, Services and Network are snapshots, not live views: each says when it
was taken, and Refresh asks the computer for a new one on its next check-in.
They have no buttons that change the computer — restarting a service is a signed
maintenance action on Actions, and ending a process is not offered
(diagnostics.md). Old links
(?tab=notes, ?tab=timeline, ?tab=access-review) open History.
The Performance tab is described in metric-history.md, with its screenshot; Processes, Services and Network in diagnostics.md.
- Patches — the tab.
- Patch status — the evidence (which agent's scan, how old), restart pending, and who manages patching where the signed agent can tell.
- Install updates — deploying from Rem0te; here it says why it is not available on this computer.
- Scan now — asks the computer for a new Windows Update scan on its next check-in.
- Offered — the updates the last scan listed, with KB, severity and size.
Use Connect on a computer's card in Devices → My devices, on its row in All devices, or at the top of its page.
- My devices — the computers this account may connect to.
- Connect — hands the machine's ID and password to your local RustDesk client and starts the session.
You do not need to type a password. Rem0te holds the credential for each machine and passes it to RustDesk for you; if you are prompted for one, something is wrong — see Troubleshooting below.
Requirements on your own workstation:
- RustDesk must be installed.
- Your RustDesk must be configured for this server. If it is pointed at the public RustDesk network it will look up the ID there and either fail to find it or reach a completely unrelated machine.
Every connection is written to the audit log with who connected, to what, and when.
Managed Secure Access is off on every production server; this section applies only where an administrator has switched it on. It has been tested only in a Windows Server lab, never on a desktop or with a real connection.
A computer whose page says Managed and verified has no standing password. Pressing Connect (or View only) asks the computer's agent to open one session for you:
- The dialog shows progress: Authorized, Received by the agent, Applied. It waits up to 60 seconds. If the agent does not answer (the computer is offline, or busy), the session is cancelled and the dialog says why; nothing was handed over. Try again later — there is no fallback to an old password.
- When it says Applied, press Download Connect script and run it. The script carries a one-time password for this session only; it works once.
- What you may do is decided by your role or preset, not by the button: a view-only preset gets a view-only session, and if your business requires consent the person at the computer is asked to accept. Clipboard and file transfer are on only when your business allows them. A TCP tunnel and a remote restart are never available on a managed computer.
- The session ends when you disconnect, at your session limit, after 15 idle minutes, or when someone disconnects it from the console.
The computer's page lists its managed sessions with a timeline. Launched means the Connect script was downloaded; it is not Connected, which appears only when the agent saw RustDesk accept a connection. The agent cannot see who connected or whether consent was accepted, and the timeline says so.
You can end your own session with Disconnect. A business's Owner or Admin, and Platform Admins, can disconnect any session there, all of a technician's, or all of a computer's (Devices → Managed sessions). A containment pause on remote access ends open sessions too.
Use Remote Support → Quick Connect for one-off help where nothing should be installed permanently — a machine you do not manage, or a person you are helping once.
- Remote ID — the 9-digit number their Quick Connect client shows.
- Password — the short code shown next to it.
- Who are you helping — optional label for the session record.
Send them to the public download page (shown on the right of that screen), they run the client, and they read you the two values. Nothing is installed as a service and no managed computer is created — closing the client ends their availability.
Rem0te does not store the Quick Connect password and never displays it again.
Devices → Add device (the Add device link under Devices in the menu, from 0.64; the button on the Devices page; or the Add device tab) walks through it in six steps on one page:
- Business — only the businesses you manage are offered.
- Site — or No site; only that business's sites.
- Who may connect — specific people, or everyone in the business.
- Continue — the three choices are fixed into the link when it is made.
- Wait for the device — like every later step, greyed until it is its turn; it is where this page watches the new device arrive (step 4, below).
- Step 1, done — what was chosen.
- Operating system — Windows, Linux or macOS, each saying which agent it installs.
- The step-up note — making the link asks you to confirm it is you.
- Make the install command — makes the single-use link.
-
Run the installer — shown once. (The token is replaced with a placeholder in this screenshot, which predates 0.64: from 0.64 this step has two tabs, Download installer and Command, described below.)
-
Wait for the device — checks every few seconds; Revoke this link cancels an unused one.
-
First health — fills in from what the server actually receives once the device checks in.
-
Business and site. Pick the business, then a site or No site, and who may connect — specific people, or everyone in the business. All three are fixed into the link when it is made: the machine that runs it cannot pick a different business or site. You can only pick a business you manage; the server refuses any other, and a site that is not that business's.
-
Operating system. Windows gets RustDesk and the Rem0te agent (heartbeat, inventory, monitoring, update scans). Linux (0.32) and macOS (0.33, run with
sudo) get RustDesk and the read-only agent: heartbeat, inventory and pending updates, and nothing the console sends is acted on (inventory.md). -
Run the installer. Two tabs (0.64). Download installer is a button: on Windows Download .exe, a one-click installer (run it, accept the administrator prompt), with the same link to copy and send to the person at the machine; on Linux and macOS the install script as a file, run with
sudo bash install-rustdesk.sh(orinstall-rustdesk-macos.sh). Command is the one-line command (PowerShell as Administrator, or a root shell). Windows also has Download .msi (0.65) and the silent linemsiexec /i rem0te.msi TOKEN=… /qnto copy; the .msi needs the server to publish the signed agent, says so when it cannot be built, leaves the agent in place when it is uninstalled, and has never been run on real Windows. There is no .pkg; for many computers at once see deployment.md. Every choice carries the same single-use enrolment token that expires in 24 hours; downloading does not use it up, the computer's first check-in does. It is shown once: reload the page and only the waiting continues. If the command is lost, revoke the link and make a new one. Making it asks you to confirm it is you (password, authenticator code or passkey) unless you did in the last 10 minutes (0.33): a link adds a computer to the business, so a session left open must not be enough. Your choices are kept while you confirm, and if you cancel nothing is made — press the button again. -
Wait for the device. The page checks every few seconds and says enrolled only when a device has used this link. You can revoke an unused link here.
-
First health — what the server has actually received: the business and site it landed in, online, the agent and its version (after its first heartbeat, which the installer starts as it finishes), whether its device identity is bound, and whether its first inventory has arrived — or, until it has, whether it has been asked for yet, and the agent's reason if the pass failed. Each line says waiting until it is true; nothing is assumed.
-
Next — open the device, its Monitoring tab, or the business's monitoring policies, with the monitors that already apply to it listed.
A Business User who adds a machine for specific people that does not include themselves is told it enrolled, but not shown it: they could not see it in the list either.
The installer is safe to re-run; it reconfigures an existing install rather than duplicating it. To re-enrol a machine whose identity is not bound, use Make an enrollment link for this computer on its page: it opens Add device bound to that computer.
It will refuse to report success if the machine never reaches the server. If you see an error instead of a summary, read it — it prints the service state, whether ports 443 and 21116 are reachable, and RustDesk's own log. Those are the facts that distinguish a firewall problem from a service that is not running.
Already running RustDesk, and only want it listed? Register it by RustDesk ID (linked at the bottom of Add device) adds it without installing anything. It then has no agent: no heartbeat, inventory or monitoring.
- Remote Support — Sessions is its first tab.
- Connected — open sessions, then the enrolled clients currently online.
- Session History — past sessions with duration and disposition.
- Connect — start a session with that machine.
- Quick Connect — for a machine that is not enrolled; a session can no longer be recorded by hand.
Each session's status says what is actually known about it. Client opened means RustDesk started on your computer, not that you connected. Connection confirmed comes from the remote computer's agent, or from you. A launch that gets no further signal reads Status unknown after 30 minutes rather than pretending it failed or succeeded, and duration is shown only when it was measured. See connecting.md.
Updates covers two separate things: Rem0te itself, and the RustDesk client installed on each managed computer.
- Settings → Updates — Rem0te's own version, then the RustDesk server and clients.
- Update all outdated — queues every machine that is behind.
- Check — re-reads the latest RustDesk release.
Per-machine status reads:
| Status | Meaning |
|---|---|
| Current | On the latest release. |
| Outdated | Behind; an update can be queued. |
| Unknown | The machine has not reported a version yet. Endpoints enrolled before v0.8.2 only start reporting after their next installer run. |
| Queued | An update is staged and applies on the next check-in. |
Queued updates apply on the machine's next heartbeat, within about 3 minutes. An offline machine applies it whenever it next checks in — nothing is lost. A queued update clears only once the machine reports the target version, so a failed install retries instead of silently disappearing.
Customers can ask for help through their business's support link, without an
account. Requests appear under Remote Support → Support requests if you hold
support:requests.
- Claim a request to take it. If someone beat you to it, you are told who.
- Check Matched computers: computers in the business that match what the customer wrote, limited to ones you can connect to.
- Connect through the computer's Connect button or Quick Connect.
- Complete it, or Hand off to a colleague, or Release it back.
The customer's status page shows your first name and the status, nothing else. See support-requests.md.
If you have technician assignments, the sidebar shows a business switcher. Pick the business you are working in; everything you see is then that business, limited to what your assignment allows. All assigned gives an overview across all of them, and lists access that is not active right now and why (expired, outside its hours, awaiting approval). Ask for access to another business from there. See technicians.md.
Start with Test connection. Every computer's page has it, on the Diagnostics tab (the Diagnose button opens it). It checks the server, the computer and, with Check my computer, your own RustDesk, and says which one is wrong. See connecting.md.
A computer shows Offline but the person says it is on. Online is driven by the agent's 3-minute heartbeat. A machine RustDesk can reach with a broken heartbeat shows amber Reachable (RustDesk) · agent silent instead; one neither can see is Offline. A machine can be running fine while the heartbeat task is broken. Re-running the installer reinstalls the task and will now tell you if the machine cannot reach the server.
Connect says "wrong password". The credential Rem0te holds and the one the machine has are out of step, or your own RustDesk resolved the ID on the public network and reached a different machine. Check that your workstation's RustDesk points at this server first — that is the more common cause and the more serious one.
Connect says "ID does not exist". The machine is not registered with this server's rendezvous service. It is either genuinely offline, or its RustDesk cannot reach the server even though the agent can. Re-run the installer on it; the verification step now reports exactly which.
A machine was reported "installed successfully" but never appears. That was possible before v0.8.2 and is not any more — the installer now refuses to claim success unless the server confirms it has seen the machine.














