Anevry Pilots OpenFrame: From DIY Automation to Built-In AI

Dave Haertel is a partner at Anevry, a two-person MSP split between South Carolina and Massachusetts. Dave builds the technical stack and his partner brings in the clients. At the interview in January 2026 they managed 10 or 11 clients and about 75 endpoints, including five or six small townships in Massachusetts.

Before OpenFrame, Dave was building his own automation. He wired up n8n workflows and leaned on Claude to write the code. It worked, but he was the one maintaining all that code. When he looked at OpenFrame, most of his to-do list was already there: script deployment, agent updates and an AI chat agent.

This is a pilot story. Anevry put OpenFrame on three or four endpoints, next to the RMM it already runs. The agent rollout went without a hitch, scripts and updates ran clean, and the two agents haven't clashed. Dave is impressed and would love to switch, once a few more pieces are in place.

Anevry's stack was still a work in progress. The partners hadn't settled on backup, ticketing or security tools yet, and by Dave's own description they both analyze everything to death. He calls it analysis paralysis. And Dave is the one person on the technical side.

So he started automating on his own. He built n8n workflows around the scripts in his existing RMM and used Claude's code model to write what he couldn't. It made his life a lot easier. It also made him the owner of a pile of code. "I'm not a coder," he said, and keeping that code running, and fixing it when something went wrong, was all on him. He wanted something that was already built.

When his partner found OpenFrame, Dave was a little reluctant. He'd already put time into his own version. Then he saw the overlap. Deploying scripts, pushing updates and an AI chat agent were all on his list of things to automate, and the AI chat agent was right at the top.

Anevry didn't rip anything out. Dave added OpenFrame to three or four endpoints and left the rest of the roughly 75 on the tools already in production. He didn't want to experiment on live client machines more than he had to.

Rollout was a single PowerShell run. Dave pushed the OpenFrame agent out through the PowerShell tool in his existing RMM: "I ran the PowerShell from there. That part was excellent." Since then he's run the agent update command when needed and done some basic scripting, with no problems. Both agents sit on the same endpoints. He wasn't sure beforehand whether they'd conflict. So far they haven't.

Then there's the AI. Dave had looked at Mingo, OpenFrame's technician AI agent, closely enough to ask how its approval step works. Asked which feature had made the biggest difference, he first said none yet, because OpenFrame already did what he asked of it. Then he took that back and picked the agent. He'd been trying to build the same thing himself, and his version wasn't close.

This is an early read. At the interview in January 2026 the pilot ran on a handful of endpoints, and Dave said he still needed more time with it, so there are no hours-saved numbers here. What he could report is that nothing went wrong. The agent rollout was "very smooth," updates and scripts ran without trouble, and OpenFrame has run beside the existing RMM with no conflicts. He called the devices side "great" and "very solid."

The second finding is about build versus buy. Script deployment, agent updates and an AI chat agent were on his own automation list, and he found all three already in OpenFrame. On the agent, he said his own n8n and Claude version is "not anywhere near as sophisticated" as what OpenFrame has built. Asked whether OpenFrame would affect his ability to take on new clients, he answered "in a positive way, yes" and said he saw no negative impact.

What he still wants reads as polish and missing pieces: a PSA with ticketing, including email commands to update tickets; image-level and Microsoft 365 or Google Workspace backup; a knowledge base the AI fills in from tickets; a Linux agent for a couple of devices; EDR integrations; letting the AI handle low-risk tasks off-hours without waiting for approval; a Slack alert when an AI session needs approval; and remote desktop fixes, so the first click lands in the login box and the window can go full screen.

His take: "Overall I'm, I'm impressed. I like it. I think you guys are on the right track for sure." He'd love to move Anevry over and run one platform instead of a stack of commercial tools. Not yet, though: "we're not going to be able to do it until we get a few more pieces together."