Adam ListekWith Remote Desktop Manager 2026.2, the AI assistant takes a meaningful step forward toward being a practical operator feature you can lean on day to day.
At Ignite in late 2025, we showed what was possible by driving RDM through Visual Studio Code and RDM’s local MCP server, a powerful proof of concept that demonstrated an assistant that can understand intent and execute real work in live sessions.
In 2026.2, that capability becomes even easier to use because the workflow is simply in RDM: the assistant understands which sessions are open and can carry out the steps within them. Fewer tool hops, less “copy this command, paste it there,” and more completed tasks.
Here is what that looks like in practice:
- One place to work: prompts, context, and actions stay in RDM, rather than bouncing between tools.
- Multi-session execution: a single request can span Windows RDP, Linux SSH, and more.
- Governed automation: approvals are built in, so you can move fast without giving up control.
Under the hood, RDM exposes structured capabilities through its built-in Model Context Protocol (MCP) server. The model decides what to do, and RDM executes it against your RDP/SSH/Agent sessions, with governance controls layered in.
Where it pays off day to day:
- Setup and tooling: install what you need across machines without rebuilding the same command sequences.
- Maintenance: apply updates, then leave a clean record you can paste into a ticket or handoff.
- Triage: pull the top issues from logs quickly, so you spend time fixing instead of filtering.
Three real-world wins with the AI assistant
Install across Windows and Linux without switching tools.
One prompt installs Sysinternals Suite via WinGet in Windows, installs btop via APT in Ubuntu, and launches the right tools. The point is not the packages; it is the pattern: one intent, two sessions, and the usual fiddly details (like the Sysinternals EULA) handled during the run.
Patch, then produce a record.
Routine updates are rarely “run the command and move on.” This clip updates 7-Zip, checks for other pending tasks, and writes a clean summary to Notepad before jumping to Linux to update jq. That “do the work, then leave a readable trail” is what makes the assistant useful in real operations.
Quick triage, not log spelunking.
The last clip asks for the top issues on a Windows host (Event Viewer) and an Ubuntu host (system logs). The assistant iterates when the first query does not return a signal, then returns a ranked summary and concrete next steps. It is the kind of morning health check or first-response pass that usually eats up time across multiple consoles.
Governance that matches how teams operate
The approvals mode you see in the videos is intentional. Most teams will keep Default approvals on for change control. When you are in a lab, running a scoped maintenance task, or recording a demo, Bypass approvals can make sense when you want an uninterrupted run. The important part is that this choice is now built into the workflow rather than bolted on.
Faster IT operations with AI in RDM
RDM 2026.2 makes the AI assistant into an operator feature inside the tool that IT admins already use for sessions and access. Routine installs, updates, and first-pass triage can go from intent to executed work (with governance) without switching consoles.
Download the latest version of Remote Desktop Manager and try the AI assistant in one of your day-to-day tasks. Then share feedback, ideas, use cases, and any bugs you run into in the Devolutions forum.
Steven Lafortune
