Rezolve.ai launches Rapport, an Agentic Endpoint Product That Diagnoses and Fixes Device Problems Under Human Approval

One agent replaces monitoring, management, patching and asset tools, adding an AI technician for unscripted faults.

Rezolve.ai announced Rapport, an agentic endpoint product that investigates a device problem, explains what it found, and carries out the fix once a person approves it, without a remote support session and without an employee having to stop work.

IT teams typically run four or five separate agents on every laptop: one for experience monitoring, one for remote support, one for patching, one for asset discovery. Each carries its own console, its own contract and its own update cycle. Rapport covers all four in a single agent, and adds a capability those tools typically lack.

Also Read: AiThority Interview with Michael Kwok, Vice President, IBM Bob, and Canada Lab Director

Existing automation typically resolves only the problems somebody anticipated in advance, by running a script written for that fault. Rapport composes its own checks while it investigates, which means it can work a problem that has never been cataloged: the long tail of device faults that today reaches a senior engineer or a remote session.

The product is built around a control model rather than autonomy. Read-only checks can run automatically where policy allows. Any command that changes a device, or needs elevation, waits for a named approver. Destructive and disallowed commands are refused outright. Every session is recorded and exportable as audit evidence.

“Most tools in this category were built to tell you something is wrong,” said Saurabh Kumar, Chief Executive Officer of Rezolve.ai. “Very few were built to do anything about it. The result is dashboards that get more precise every year while the ticket still waits for a person. We built Rapport because the useful question is not how quickly you can detect a problem. It is how much of the work between detection and resolution a person still has to do.”

“The agent runs on the device and the reasoning runs in the cloud, and the agent decides what is allowed to execute,” said Udaya Bhaskar Reddy, Co-Founder and Chief Technology Officer at Rezolve.ai. “That separation matters. A model asking for a command is not permission to run it. Classification happens locally, mutating commands wait for an approver, and the whole thread is recorded. Outbound HTTPS only, with no inbound port and no firewall exception.”

“The difference shows up on the problems nobody wrote a script for,” said Joshua O’Brien, Head of Applied AI at Rezolve.ai. “A scripted tool handles the faults you already understood. Rapport reasons about evidence it gathers as it goes, then shows a technician exactly what it found and exactly what it wants to run. The technician stays in the decision, which is the only way this gets deployed across a real device fleet.”

Also Read: Why Coordination, Not Code, Is the Real Constraint on Shipping Software Today

[To share your insights with us, please write to psen@itechseries.com]