Self-hosted remote support for customer businesses, powered by RustDesk.
One operator, many customer businesses — each with its own computers, its own people and its own history. Rem0te is the portal around a self-hosted RustDesk server that keeps them organised and separate.
Quick start · How access is protected · Screenshots · Walkthrough · Documentation · Security · MSPReboot
The operator's view — every customer business, computer and session in one place. Full size · 90 more screenshots
Which technician can reach this customer's PC? Only someone the server says yes to, for that one computer, at that moment. Hiding a button is not how it works.
Take two customers, Acme and Birch, that both have a PC named FRONTDESK:
- The right machine is chosen by its id, never its name. Connect, the password, the launcher token and the session all name one computer's database id. The list a person picks from contains only computers of the business they are working in, re-read from the database on every request. A technician assigned to Acme sees one FRONTDESK.
- Without access, nothing is released. If that technician opens Birch's FRONTDESK by URL or API, they get the same 404 Computer not found as for an id that does not exist, and the attempt is audited.
- Within Acme, it depends who is asking.
- A Business User needs
computers:connectand a grant for that PC. - An L1/L2/L3 preset technician never receives the password. Interactive control needs a temporary grant that someone else approved.
- On endpoints as shipped, preset launches are refused, because the computer cannot enforce consent or view-only. The portal does not open something it cannot restrict.
- A Business User needs
- Enrollment is bound to one business. The installer link carries a single-use, expiring token minted in that business. Every heartbeat authenticates with a per-device secret, not the RustDesk ID, and nothing a device reports can move it to another business.
What evidence will I have afterwards?
- Audit entries for every password reveal, launch grant, refusal, launcher token and session change.
- A session record that says how each step is known: Client opened is not a connection; Connection confirmed comes from the endpoint's agent, or from the technician, recorded as theirs.
- A per-person list of the passwords they have seen.
What you will not have. Rem0te authorizes the launch. RustDesk carries the session. So the portal cannot see what happens inside a session or end it, and anyone holding a computer's password can connect without the portal. That is why offboarding tells you which passwords to rotate.
Clipboard, file transfer and view-only for business members are RustDesk defaults, not Rem0te restrictions. So are terminal and chat, and Rem0te ships no remote shell of its own. Each control, the layer that enforces it, its default and its limits: Settings → Security → Security Features, and docs/security-model.md.
Found a problem? Report it privately: SECURITY.md.
Every change on a computer below (services, processes, installing and removing software) needs the full (signed) Windows agent and an approval signer, and none of those changes has yet run on a real Windows computer: they have been tested against a simulated one. Ask AI and Create ticket have been tested only against stand-ins, never a real provider or PSA. What each agent can do says what is proven and what is not.
- HP and Apple warranty lookups, and why a warranty is empty (0.61). Settings → Warranty can now look up HP Inc. computers (HP's Warranty API, after HP approves your developers.hp.com enrolment) and Macs in your Apple Business Manager, beside Dell and Lenovo. Only the serial is sent. Every computer with no warranty date now says why — vendor has no lookup (HPE, Surface, ASUS, Acer), lookup not switched on, serial missing, no record, or the failure — on its card, in the report and on the dashboard. The HP and Apple lookups have been tested only against a stand-in. Hardware warranty
- Restart windows (0.60). "Restart if Windows needs it, nightly 02:00–04:00" on a business, a device group or one computer, which can also say No automatic restarts; the computer's own setting wins, then a group's, then the business's. A window is approved once on the approval signer (or is in force at once in direct mode), and each restart is its own signed plan with the 15-minute warning, deferred while a technician is connected through RustDesk and tried again 15 minutes later and in the next window. The Patches tab says whether, and when, the computer will restart. With nothing set, nothing restarts a computer automatically. Restart windows
- Groundwork for trusting a self-hosted Keycloak (0.59). Each sign-in provider now has trust settings that
start off: whether its MFA counts as Rem0te's second factor (from
acr, checked againstamr), whether its passkeys count as a passkey, whether Platform Admins may use it (only with a passkey there), an organization it must name, andprivate_key_jwtwith a key kept sealed here. Signing out at the provider ends the session here (back-channel logout), and signing out here ends it there. SSO-only accounts can confirm it is them through the provider. A business can be set to SSO required once everyone in it is linked, with up to two break-glass admin accounts that can always sign in, and the login lockout now survives a restart. Tested against a mock provider only. sso-keycloak.md · break-glass.md - Automation schedules (0.58). An approved library script or a software catalog package can run every few hours, daily, weekly or monthly, on up to 25 named computers or on a device group with a cap (Automation → Schedules). Two passkey approvers approve the schedule once on the approval signer, never its author or the script's writers. The signer then signs each run and holds it to the same task, computers, window and offline rule (run when online, skip, or expire after N hours, signed into each plan). Runs never overlap. Pause is immediate; revoking the script, archiving the package, a containment pause or revoking the schedule stops every later run; a change is a revision approved again. A single script run or install can also be scheduled once. The signer must be reinstalled from 0.58; the agent is unchanged. Automation schedules
- Deploying agents and adopting computers (0.57). A Platform Admin makes a multi-use deployment key for a
business (at most 90 days and 10,000 uses, shown once, stored as a hash) on Devices → Deploy agents, which
also gives copy-paste instructions for a Group Policy startup script, Intune, an existing RMM, Linux and macOS,
with the key filled in by the browser. The existing installers take it in place of a single-use code. By default a
computer that enrols with a key waits under Devices → Unassigned, with its identification, hardware, source
address and how it arrived. Nobody in a business can see it, and every remote and action path refuses it
(
ENDPOINT_UNASSIGNED), until a Platform Admin assigns it, one or many at a time, to a business, site and groups, with a fresh sign-in and an audit entry per computer. Assignment rules (deployment key, domain, address range, hostname prefix) only suggest a business, never assign one, and never use who is signed in. deployment.md - Hardware warranty and support links (0.56). Each computer's page links to its manufacturer's support, warranty and driver pages (Dell, Lenovo, HP, HPE, Surface, Apple, ASUS, Acer; virtual machines say they have no hardware warranty). Warranty dates can be entered by hand, or looked up from Dell or Lenovo once a Platform Admin switches that on (off by default; it sends the serial number). They feed a device-list column and filter, a Warranty report, a dashboard tile and an opt-in Warranty expiring monitor. The lookups have been tested only against a stand-in. Hardware warranty
- Alerts for drives, uptime and updates, and automatic PSA tickets (0.55). Four new monitors read what the server already stores, with no agent update: every fixed drive (one alert per drive), a computer not restarted for N days, security or critical updates outstanding for N days (signed agent's patch scan), and a failed update install. Three of them join Add recommended. A rule under Settings → Monitoring opens a PSA ticket when an alert of at least a chosen severity opens. It sends once per alert, after the alert's hold, never in a maintenance window or under a notifications pause, inside a per-business hourly budget, and retries a failed send three times. The dashboard has a Recent alerts card with each alert's ticket reference.
- Restart scheduling and restart windows (0.54). A restart can be asked for now or at a date and time up to 14 days ahead, in a named zone, from a computer's Actions tab, the Patches tab's restart banner, or the device list for a selection or a device group (at most 200, one approved plan each). Only if a restart is required leaves the decision to the computer, which reads Windows' own pending-restart state. A patch policy, or a company-wide one, can now have a restart window — restart if required or a maintenance restart — and the approval signer signs each restart only inside it. A signed-in user never stops a restart: they see the 15-minute warning, and unsaved work is lost. Only a technician connected through RustDesk does, unless the request (or the policy) carries the new, never-ticked-by-default exception Restart even if a technician is connected through RustDesk. Needs the 0.54 agent (a signed publish) and an updated approval signer.
- Monthly patch windows (0.53). A patch policy, or a company-wide one, can install on the nth or last weekday of each month plus 0–14 days — second Tuesday + 2 is the Thursday after Patch Tuesday — or on a day of the month from 1 to 28. The form shows the next three openings in the policy's time zone, worked out by the same code the approval signer holds every ring to. Weekly policies are unchanged. The approval signer needs updating to 0.53 before it accepts a monthly policy; the agent does not.
- See where every task is, and schedule update installs (0.52). Every action, install, restart, script, snapshot, scan and agent upgrade you send shows its steps — scheduled, waiting for approval, queued for the next check-in, handed out, running, done — with per-update results, on the computer page, on Automation → Tasks and in a toast when yours finishes. After an install that needs a restart the Patches tab says so and offers Schedule restart. Install updates and a deployment's next ring can start at a date and time in a named zone, signed into the plan; scheduling has been tested against a simulated agent only. Staged patching
- One computer, one agent; a tray icon you can find (0.51). A computer that runs the signed agent stays signed even when the standard agent's leftovers still check in: those reports are ignored for everything that acts, and the signed agent removes the leftovers by exact name. The tray icon is a speech bubble inside a bright ring whose colour gives the state, and Open a support request appears only when a PSA connector is set up, becoming a ticket there. Neither the clean-up nor the icon has run on real Windows yet. From 0.53.1 the menu starts with who it is for — the business, the provider and its support phone and e-mail — and the tooltip and chat window use the same names instead of fixed text; the menu has not been photographed on Windows. Capabilities
- Patch policies for a whole business, changed per computer (0.50). A company-wide patch policy covers every Windows computer of a business on the signed agent: the portal picks the canary, puts the rest in rings of at most 25, and lists any computer it left out with the reason. Each computer's Patch policy is Inherit, Exclude or Use policy X (for example one for servers), and its Patches tab shows which applies and why. When computers change, the portal proposes a new revision, approved like any policy: after a debounce, at most once an hour per business, never during a deployment. Staged patching
- Choose which changes ask "confirm it is you" (0.49). Settings → Confirm it is you lists 30 functions in nine groups, each Ask or Don't ask. Your own password, authenticator and passkeys, and turning a confirmation off, always ask. Signing in
- Reinstall a computer that is already here (0.48.2). A download link bound to one existing computer, from its page or its row in Devices, for an agent that stopped or was removed. It re-attaches to the same record. Troubleshooting
- Ask AI and Create ticket (0.48). Every event-log entry and alert has Ask AI, which explains it in plain words after redacting names, addresses, accounts and serials, and Create ticket, which sends it to your PSA by email-to-ticket, a signed webhook or ClientSt0r. You bring your own Anthropic, OpenAI or compatible key, kept encrypted and never shown again. AI and tickets
- Services and processes (0.45). Start, stop or restart a service, change its startup type, or end a process, from the Services and Processes tabs. Each is a signed action; protected Windows, security, Rem0te and RustDesk services and processes are refused on both sides. The Actions tab shows the Print Spooler's state (0.47.1). Approved actions
- Software from winget (0.41, 0.43). Search or browse winget's catalog and add packages, up to 25 at once, each pinned to a version and installer hash. Remove an installed program from the Software tab. Software deployment
- Chat with files and an inbox (0.44). Emoji, pictures and files in chat, and unread replies in the header from anywhere in the console. Chat
- Automatic agent updates per computer and per business (0.46), and fewer "confirm it is you" prompts for everyday changes (0.47, per function from 0.49). Updates
- Direct mode for a single operator (0.40): requests are signed at once, with no second approver.
- The Devices list fits the window (0.47.1, 0.48.1): from 1280 to 2560 pixels with the sidebar open. Last seen, Agent, OS, Site and Logged-in user each appear at their own width (1440, 1536, 1700, 1860 and 2000 pixels).
Patch status you can trust the age of. Each computer's Patches tab now leads with its evidence: which agent reported it, when, and whether that is fresh, stale or unknown. Then last boot, restart pending, the relevant KBs, and who manages patching — WSUS, Intune and the rest, with what each means, or unknown from the standard agent, never "nobody". Each update from an approved plan shows whether a later scan confirmed it. A restart pending alert sent to ClientSt0r carries a redacted copy of the summary for the ticket. The ticket ↔ RMM contract is published with signed examples from a real run, and capabilities now name each organization's business and pause state. The whole workflow is proved end to end against a reference peer. Against the real ClientSt0r it is measured not to work yet. capabilities.md · CLIENTST0R.md
Know when a computer needs you. Monitors watch for a computer offline, a disk nearly full, the processor or memory pinned, a Windows service stopped, matching events in an event log and a restart pending for days. They are inherited global → business → site → computer. One alert per problem: repeats are counted, and a site that loses three or more computers raises one site outage instead of a page per machine. Maintenance windows hold notifications for planned work, and are not containment. Up to four escalation levels notify webhooks, and alerts can be offered to ClientSt0r, which links back the ticket it opens. Every computer gains installed software and a change history. Settings → API keys is a new screen (keys could only be made through the API before). The documentation is reorganised: a grouped index, a contents panel on every page, search with highlighting (press /), and annotated screenshots of the main screens. monitoring.md
Stop writes during an incident, in three places. A global write pause, or a suspension of one business, stops
approved actions (including plans already approved), remote connections, agent updates and ClientSt0r jobs. A
Business Owner can suspend their own business, and the MSP can't lift that. An MSP pause is lifted by a second
person, always with a fresh sign-in. The approval signer keeps its own copy, which only its operator can lift.
A signed control/pause.json, published from the release machine, makes the agents refuse even if the portal is
compromised. The audit log is tamper-evident. Entries are hash-chained, a database trigger guards them, and
the chain is verified hourly. It can be streamed off the host with signed checkpoints and checked against that
copy with rem0te-audit verify. /api/v1/healthz and rem0te-external-check.py let a monitor on another
machine watch it. containment.md · incident-recovery.md
Windows updates in stages, a canary first. A deployment names Microsoft security, critical, rollup and ordinary updates and the computers in each ring. Ring 0 is one canary computer. Each ring is one plan that two people on the approval roster approve with a passkey, and the next ring can't be requested until the canary reports every update effective: no longer offered, recorded in Windows' history, no restart pending. Too many failures and the deployment pauses itself. Each computer asks Windows Update itself what it may install, and refuses to install anything if WSUS, Intune or Configuration Manager manages it. It never restarts. A Patches tab on every computer shows what Windows Update offers it. staged-patching.md
Fix one thing without connecting, after people approve it. The first approved action starts a Windows computer's Print Spooler, once. Approvers authorize the exact plan with a passkey, on a separate approval page that the portal neither serves nor controls. Servers need two approvers, and a business can require its own sign-off. The computer re-checks the signature, its own identity, the window and the approvals before it runs the job, at most once. A result reports the service state, never that a document printed. ClientSt0r can propose an action but never approve one. approved-actions.md
Set the server's address in the app. Every install is reached at its own address, and support links, sign-in redirects, passkeys, launcher links and installer downloads are built from it. It used to live only in the server's environment file; now a Platform Admin sets it in Settings → General, and the server checks the address really leads back to it before saving. setup.md
Find out why a connection fails, before anyone calls. Test connection checks the server, the computer and your own RustDesk, says which one is wrong, and labels every fix that would change something. connecting.md
Session states that mean what they say. "Client opened" is not "connected"; a connection is confirmed by the remote computer's agent or by you; anything unproven reads Status unknown, and duration appears only when it was measured.
MSP technicians, scoped. Give a technician (or a group) one business, some capabilities, optionally some computers, optionally for a week or weekdays 9–5, with Business Owner approval where wanted. A business switcher and an All assigned overview. technicians.md
Customers can ask for help. A per-business support link, a status page that reveals nothing else, and a queue technicians claim, hand off and close. support-requests.md
Controlled RustDesk updates. Exact versions, pilot batches, maintenance windows, deferral during sessions, bounded retries, automatic pause on failures, and a supported-version pin. updates.md
Access review and offboarding that tell the truth: who can reach a computer and why, and when someone leaves, which RustDesk passwords they had — because revoking access does not change a password already seen. offboarding.md
SSO and passkeys. OIDC sign-in matched by issuer and subject (never email), passkeys for sign-in and step-up, step-up for sensitive actions, and a recovery script. sign-in.md
Every computer describes itself. Hardware, storage, network adapters, who is signed in, uptime, and how many Windows updates it is waiting on — collected by the managed agent on its heartbeat, with live CPU, memory and disk gauges that say how old their sample is. inventory.md
Read a Windows event log without connecting. Pick a log, a window and the levels; the request is collected on the computer's next heartbeat. Five logs, allowlisted twice, behind its own permission — somebody else's Security log is not the same as a name and an online dot.
Works from Tactical RMM. Right-click an agent in TRMM and connect through Rem0te, or open that computer's page. Hostnames collide between customers, so Rem0te asks rather than guessing — and remembers your answer. tactical-rmm.md
Upgrade the agent from the console. Reinstall agent re-runs the installer on the next heartbeat, keeping the machine's configuration, password and enrolment. It is also the only way to reach a machine whose RustDesk client is already current.
A dashboard that fits one screen, with host CPU, memory, disk and live bandwidth for the operator, and a count of sessions actually relaying through the server.
Full detail in the changelog.
Every customer's computers in one place. No spreadsheet of RustDesk IDs, no "which machine was that again?". Computers are grouped by the business that owns them, with platform, online state and last-seen time.
Connect without passing passwords around. A technician clicks Connect; Rem0te authorises the request and hands the credentials to the RustDesk client on their own PC. Nothing is typed, emailed or kept in a shared note — and every reveal is written to the audit log.
Give each person exactly the access they need. A Business Owner decides who may see which computers and what they may do, without the operator being in the loop for every change.
Diagnose before you interrupt anyone. Each computer reports its own hardware, disk space, signed-in user, uptime and pending Windows updates, and you can pull a slice of its Windows event log without starting a session. Half of "can you take a look at my PC?" is answerable without touching the machine — and reading event logs is a separate permission from seeing the computer, because someone else's Security log is not the same as a name and an online dot.
→ See all 91 screenshots — the dashboard, search, the device list and every device tab, Ask AI and Create ticket, chat, task progress and the Tasks page, the latest install and scheduling one, reports, the script library, the winget catalog, patching and monthly windows, administration and configuration, in light and dark and at phone width. All captured at v0.53.2.
A captioned tour of every feature, recorded at v0.53: the dashboard, the Devices list, one computer's workspace tab by tab (agent auto-update, software, services, processes, event logs with Ask AI and Create ticket, patches with the latest install, its restart and a scheduled install, each computer's Patch policy, actions, diagnostics, chat), Automation → Tasks and the toast when a task finishes, Connect, the reinstall link, Add device, company-wide patch policies, monthly patch windows, scheduled deployment rings, the software catalog, scripts, monitoring, reports, businesses, Confirm it is you and the other settings, the audit log, and the built-in docs.
▶ Watch the walkthrough (14 min 35 s) · 81-second highlight · transcript and chapters · captions
Everything shown is an isolated demo environment: fictitious businesses and people, and computers that report through a simulated agent. None of the computer actions in it ran on a real Windows PC, no remote session is made, and the AI provider and PSA are loopback stand-ins. The walkthrough is caption-led: there is no narration and no audio track.
flowchart LR
Tech["Technician<br/>(web browser)"]
Portal["<b>Rem0te portal</b><br/>Next.js + NestJS + PostgreSQL"]
Local["RustDesk client<br/>on the technician's PC"]
RD["<b>RustDesk server</b><br/>hbbs + hbbr (self-hosted)"]
Dev["Customer device<br/>enrolled RustDesk client"]
Tech -->|"1 - sign in, pick a computer"| Portal
Portal -->|"2 - authorise, return ID + password"| Tech
Tech -->|"3 - rustdesk:// deep link"| Local
Local <-->|"4 - remote desktop session"| RD
RD <--> Dev
Portal -.->|"enrollment token + server config"| Dev
style Portal fill:#dbeafe,stroke:#2563eb
style RD fill:#dcfce7,stroke:#16a34a
Rem0te adds the portal: businesses, accounts and permissions, managed device
enrollment, credential storage and release, session records and the audit log.
RustDesk provides the remote desktop itself — the hbbs rendezvous server,
the hbbr relay, and the clients. Rem0te never sees the pixels or the keystrokes.
Browser launching, not browser-based control. Clicking Connect opens a
rustdesk:// deep link, which the OS hands to the RustDesk client already
installed on the technician's machine. A local RustDesk client is required.
There is no in-browser remote desktop.
| Managed device | Quick Connect | |
|---|---|---|
| Set up by | Running a one-time enrollment installer | The person being helped runs a client |
| Lives in | The business's computer list, permanently | Nowhere — no computer record is created |
| Connect using | Stored credentials, released on authorisation | An ID and password they read out |
| Ends when | You remove it | They close the client |
Three levels, and no reseller hierarchy.
| Level | Scope |
|---|---|
| Platform Admin | The operator. Runs the service, creates and manages every customer business, owns platform settings. |
| Business Owner | Full control of one business — its computers, people, sessions and history. Nothing outside it. |
| Business User | Exactly the capabilities their Business Owner granted, over the computers they were given. |
A Business is the boundary that matters, and it is enforced on the server rather than by hiding buttons. Full model in docs/access-control.md.
No. Rem0te runs against the open-source RustDesk server and the standard
open-source clients. The installer fetches hbbs/hbbr from the OSS
rustdesk/rustdesk-server
releases, and every client configuration Rem0te generates sets api-server = ''
— RustDesk's own API server, a Pro component, is deliberately unused. Rem0te's
portal fills that role.
No paid component, licence key or subscription is required by this project. Rem0te is MIT-licensed. RustDesk is a separate project under its own licence.
Full instructions, including manual installation, in docs/setup.md.
git clone https://github.com/agit8or1/rem0te
cd rem0te
sudo bash deploy/scripts/install.sh your-domain.example.com admin@example.comThe installer sets up Node.js, PostgreSQL, Redis, Caddy (automatic HTTPS), fail2ban, the RustDesk server and its keypair, builds the app, runs migrations, seeds a Platform Admin and starts the systemd services. It prints the login URL and admin credentials at the end — save them immediately.
- Businesses → Add Business — create the customer.
- Enroll Computer — choose the business, who may connect, and the platform. The business is bound into the token here and the machine cannot change it.
- Generate Installer — a one-time command or a Windows installer executable.
- Run it once on the target machine, as administrator.
- The computer appears in that business's list; the users you selected see it under My Computers.
Server — Ubuntu 22.04 LTS or Debian 12, 1 GB RAM minimum (2 GB recommended), 10 GB disk, a public IP. A domain is strongly recommended; automatic HTTPS depends on it. Node.js 20+, PostgreSQL, Redis and Caddy are installed for you. Deployment is plain systemd — Docker is not required.
Ports — 80 and 443 for the portal, plus RustDesk's own: 21115 (TCP), 21116 (TCP and UDP), 21117 (TCP), 21118–21119 (TCP websocket, optional).
Endpoints — enrollment scripts are generated for Windows, Linux and macOS. Windows additionally has a compiled installer executable.
RustDesk — hbbs/hbbr are installed from the latest OSS release. Generated
client configuration targets the RustDesk 1.4.x client series.
- Part of an RMM, not all of one. Rem0te monitors computers, keeps a software inventory and change history, reports Windows Update status, deploys Windows updates in approved stages, and hands alerts and ticket analysis to ClientSt0r. It runs reviewed PowerShell scripts, deploys winget software and controls services only as signed, approved actions; it has no remote shell, no third-party patching and no backup. What works depends on the agent: the standard agent reports patches read-only, every change needs the signed agent and the approval signer, and Linux and macOS get remote access only. None of the signed-agent actions has run on real Windows yet. What each agent can do has the full table.
- ClientSt0r cannot pair yet. Rem0te implements the ticket-rmm-v1 contract, and ClientSt0r 3.17.659 speaks a different protocol: four measured gaps, listed in CLIENTST0R.md.
- No in-browser remote control — a local RustDesk client is required.
- One RustDesk server per instance. Relay host and public key are platform-wide settings, not per-business.
- No trend or time-series charting. The dashboard has stat tiles, a seven-day session bar chart and a client map; there is no charting library and no historical graphing beyond that.
- Automatic HTTPS requires a domain; a bare IP works but over plain HTTP.
- Report a vulnerability: follow SECURITY.md and do not open a public issue. Private reporting is enabled on this repository.
- Who can reach a computer, and the evidence you get: docs/security-model.md.
- What is enforced: docs/SECURITY-AUDIT.md records the review passes. This is an internal review by the project, not an independent third-party audit.
- Business isolation is covered by an automated suite
(
apps/api/scripts/e2e-business-access.mjs) and static invariant checks (scripts/check-security-invariants.mjs), both run by CI on every push.
Rem0te, the RustDesk clients on endpoints, and hbbs/hbbr update
independently — docs/updates.md explains which is which. The
in-app updater is off by default and requires a GPG-signed release tag.
Start at the docs index — it routes by what you are doing.
| Guide | For |
|---|---|
| Screenshots | The full gallery — every screen, light and dark. |
| Troubleshooting | Something will not connect. Start here. |
| Technician Guide | Day-to-day use — connecting, Quick Connect, enrolling. |
| Connecting | What happens when you click Connect. |
| Clients | Every RustDesk client Rem0te hands out. |
| What each agent can do | Capabilities by agent type, configuration, and what is tested on real Windows. |
| Inventory & event logs | What a computer reports about itself, and reading its Windows event log. |
| AI and tickets | Ask AI and Create ticket on event-log entries and alerts. |
| Approved actions | Signed changes on a computer: services, processes, restarts, logs. |
| Software deployment | The winget catalog, installing, and removing programs. |
| Tactical RMM | Connecting and querying from inside Tactical RMM. |
| Setup | Installing and operating a server. |
| Architecture | What runs where, and the RustDesk config chain. |
| Access Control | The three-level model and capabilities. |
| API Reference | Every route and its required capability. |
| Public API | Scripted RMM/PSA integration, scopes and examples. |
Issues, ideas and pull requests are welcome.
Every change bumps the version and adds a changelog entry — see CLAUDE.md.
Rem0te is published by MSPReboot, an MSP consulting practice that also releases free, self-hostable tools for MSPs. The other open-source projects listed there:
| Project | What it is |
|---|---|
| ClientSt0r | Customer documentation, assets and knowledge base |
| OPNMGR | OPNsense firewall fleet management |
| Depl0y | Proxmox infrastructure and VM deployment |
| St0r | UrBackup backup visibility and administration |
These are separate projects with their own repositories and licences.
Rem0te is MIT-licensed and free to self-host — the install guide above is all you need, and nothing here requires an engagement. If you would rather not run it yourself, MSPReboot offers hosting and support for Rem0te: get in touch.
MIT — see LICENSE.
Project managed by Luna, a German Shepherd Dog with discerning taste in remote support software. 🐾











