An MSSP Director Pilots OpenFrame Before a Law Firm Rollout

Michael is Director of Cyber for three partner MSPs. He tests security products for them before they reach clients. He also runs his own MSSP with a few direct clients: entrepreneurs and single law practices with one or two lawyers. Most of his clients are law firms held to a regulatory standard.

At the interview in March 2026, he was piloting OpenFrame with those direct clients, on production Windows and Mac endpoints, not test devices. He's an early adopter, and he builds his own AI technicians. He wasn't shopping for a bargain: "I'm not in it for a commercially cheap option." He wanted a platform built with agents in mind.

For an MSSP, the bigger cost is technicians, not vendors. In Michael's words, the tools aren't good enough, so "you need level three even to do level one stuff."

He'd tried open-source tools before, mostly for Microsoft 365 configuration and drift management. That meant 80 pages of PowerShell script, and scripts that broke when a cmdlet didn't exist in the PowerShell version on a given endpoint. He found those tools finicky.

His clients raise the bar further. Most are law firms that engage him for security practices with a regulatory standard attached. So he tests new tools with a few direct clients before anything reaches the wider base.

Michael put OpenFrame on his direct clients' Windows and Mac machines. Device onboarding stood out. "I do feel it was a very easy onboarding for devices," he said. After the open-source tools, that was the part he rated most: "the positive thing is onboarding stuff to OpenFrame has been pretty good."

Windows ran fine. Mac took more work, because macOS needs screen recording and accessibility permissions granted by hand. He couldn't expect end users to manage that, so he kept a second remote-access tool alongside OpenFrame to grant them. Performance wasn't the issue, he said. Installation and permissions on Mac were.

Fae, the AI agent that sits on client machines, was "pretty good as a generic knowledge base for a Windows user." He liked the scripts too. "We're in a CLI world," he said, and scripts are where some of the newer RMM tools are moving. The ticket integration was simple. Mingo, the AI agent for technicians, was where he wanted more: it started each session from a blank slate.

At the interview in March 2026, OpenFrame was running on production Windows and Mac endpoints for his direct clients. Onboarding was very easy and performance was fine. His partner MSPs and his wider base of law firms weren't on it yet. OpenFrame was in beta then, and that was the one thing stopping a wider rollout to clients with regulatory standards to meet.

His plan was to keep testing: "I will test it now so that as soon as I can deploy it more widespread, I can."

The reason he's testing it at all is the design. "I don't like anything else where they're bolting on AI," he said. "I think if you guys are building with agents in mind, that's what I like."

What he still wants, framed as polish. Mingo should carry context between sessions, like a per-device history and the last few tickets, instead of starting from a blank slate. A tool approved once shouldn't need approving again in the same session, and approval requests should reach him in Slack or Teams. On Mac, he wants clear permission docs, and ideally Mingo turning on screen recording and accessibility from the terminal. He'd also like ticket auto-triage by urgency and keywords, backup with local, cloud and geo-sovereign options plus email archiving, and agent-friendly access, like a CLI, so his own AI technicians can work with OpenFrame's agents.