Now Available: Syncro MCP Server

Announcements & Updates
1

Hi folks! Excited to share that Syncro’s MCP Server is now generally available and included on all plans.

The MCP Server lets you connect Syncro to tools like Claude, ChatGPT, Copilot, etc., to get all sorts of work done: create and assign tickets on the go, generate reports and artifacts for QBRs and onsite visits, ask questions about different queues or billing info, and get fast answers all in the AI tools you’re already using.

Permissions are aligned with whatever the user/tech can access in Syncro, plus any additional configurations you may have in your AI tool.

Sharing a few links below with more info. If you’re in Claude you can very easily find + connect by simply searching for Syncro under Customize → Connectors.

We’ll cover more on this in our monthly release webinar and should have additional prompts and skills coming soon. Stay tuned for more!

2

like the prompt ideas - found it to be very useful so far but landed at a time where I was very time poor :frowning:

3

I have been experimenting with this today through ChatGPT. I have a Project call Client Issues 2026, and for every phone call I have with a client, I open a chat thread in this project and document the conversation, the issue and the resolution. In the past I then would have to go into Syncro and open a ticket and paste the ticket summary. Now I have been successrully having Chat create the ticket.
Each time I do this I am asked by Chat to allow the use of the 3rd party plugin. Can this be bypassed somehow?

4

Cool!

I considered using this to make a workaround for the Asset Audit Report missing its RMM Alerts, so we can do QBRs again. But I realize it can’t assemble what Syncro has already lost, so alas for my first use case.

5

thanks for sharing this! the team did some quick testing here and was unable to replicate that issue in GPT, but we’ll take a closer look. I’ll loop back if we find anything, or if the issue persists feel free to open a ticket and someone can try to help you directly :slight_smile:

6

Thanks again for the feedback. Invaluable. This maybe controlled on the ChatGPT side. Here are some controls to review. https://help.openai.com/en/articles/11487775-apps-in-chatgpt#permission-options

7

I just finished half vibe coding, half hand coding one last night that’s running in a container :grimacing: all those hours and tokens wasted. But to be fair, it also talks to UniFi, RustDesk and 365 (read only on 365 for now) I’ll def give this a go.

8

Hello,

With this MCP are we able to interact with scripts?

Meaning can I create scripts, edit existing scripts, delete existing scripts, read existing scripts including name, description, category, script variables, required files, etc?

If so that would close a gap in the website gui functionality where we cannot bulk delete scripts or bulk export scripts still.

Now I have a scraper that does the exporting regularly for me but it’s clunky and requires running code directly in chrome dev tab.

But having that tied into the MCP would be invaluable.

Then we can create documentation automatically based on what scripts do, what alerts they are throwing, what custom asset fields they read and set, what policies they are in, how they interact with other scripts, how everything is wired up and running on a regular basis so that the documentation is living.

It also lets us more easily identify what scripts are dead and can be deleted.

As a side note to that something super useful that is missing is the last time a script was executed on the platform. That one simple variable would let us know which of the hundreds of scripts could be deleted because they haven’t been run in years for example.

But any information on that? Full script access via MCP?

9

You should check this out:

I will be integrating the MCP server ASAP. I have the design and implementation plan. I have been waiting for the release of the MCP server.

It is my understanding that the MCP server will let you ask: “Find the computers have not been restarted in 3 days?”

With DrModule, you will be able say: “Restart the computers that have not been restarted in 3 days.” Then AI will get the list and execute the scripts.

At least, that’s the plan. The system already works but without the help of AI. I use a maintenance review function to identify issues and make recommendations to resolve the issue. Then a recommendations engine will execute the remedies. Functions will confirm the situation, perform the remedial actions, log the details and satisfy the recommendation(s).

10

Thank you - that looks very interesting.

With this are you able to have full interaction with existing scripts inside Syncro? Like are you able to get information about every script already in Syncro as I mentioned above? I only ask because to my knowledge the API does not expose that data so I’m not sure how this project is getting that data from outside of Syncro.

I do know the API provides script execution so I could see that mechanism working but I don’t know if script execution means only executing existing scripts by ID or something else.

11

Up to this point it has been focused on scripts running on endpoints. That has been the biggest return so far.

After a chat with copilot, I realized I already have the system build for AI. We call it the controlled execution layer for AI-assisted MSP operations.

I am working on the implementation now.

I am behind because I did not have access to the MCP until it was released yesterday.

But right now, I use the module in my production environment. I have ~400 endpoints and I processed 467 tickets last month. Most ere automated assessments, maintenance runs or remediations. I cannot tell you how much time this has saved me.

12

Real life results from a prompt in case your interested.

Prompt: Can you let me know which computers have not been rebooted at CORP in the last 14 days

Result:
Here’s what CORP’s Syncro data shows for computers not rebooted in the last 14 days (today is 12 Aug 2026, so cutoff is 29 Jul).

Of CORP’s 42 monitored computers (excluding switches, printers, UPS, router and NAS, which don’t report boot times), 18 haven’t rebooted within 14 days. I’ve split them by whether the RMM agent is actually still checking in, since a few of these look like they may be offline or decommissioned rather than just needing a reboot.

Active agents, genuinely overdue for a reboot:

CORPWKS03 (170 days), CORPHVSRV01 (123 days), CORPWKS03 aside, also CORPWKS47, CORPWKS44, CORPWKS42 (23 days each), CORPWKS40 (21 days), CORPWKS112 (20 days), CORPSRV02 (15 days), CORPWKS49 (14 days), and CORPWKS103 (125 days, agent last checked in 3 days ago).

Agent hasn’t reported recently, so the boot data may be stale (worth checking these are still in use before chasing anyone about them):

CORPWKS113 (agent silent 17 days), CORPWKS101 (agent silent 13 days), CORPWKS109 (agent silent 13 days), CORPWKS116 (agent silent 124 days), CORPWKS104 (agent silent 314 days), CORPWKS111 (agent silent 376 days), CORPWKS28 (agent silent 526 days), CORPWKS29 Ben (agent silent 517 days).

That last group of four (WKS116, WKS104, WKS111, WKS28, and Bob’s WKS29) haven’t checked in for months, so they’re either powered off long-term, retired, or the agent’s been removed — worth a quick check in Syncro before assuming they need a reboot push.

13

I see, that is interesting.

14

This is great news! I would like the ability to generate RMM agent installers through MCP - but a huge win here!

15

Yeah according to the API we do have the ability to post execute a script based on an asset ID and a corresponding script ID.

But at least from the documentation I see no get mechanism to see what scripts are available or what IDs match to names. How to know what runtime variables are needed or available (dropdown vs freeform) etc.

Other than going into Syncro manually and building your own database of script names to IDs and vars, which is tedious with no bulk exporter option.

Now you can write code to scrape it and run through your browser, but doing that there seems to be multiple IDs attached to each script. A shorter one and a longer one.

So it just seems weird to include this in the API but in such a limited fashion.

16

Very cool!

I am very excited to integrate my module. From the research I have done, it should be pretty easy to integrate.

@lance hit the nail on the head… “I see no get mechanism to see what scripts are available or what IDs match to names. How to know what runtime variables are needed or available (dropdown vs freeform) etc.”

The integration I am building will identify the scripts by name and have description so you can say: “Restart the computers that have not been restarted in 3 days.”
and it will know which function to run. The functions are written to be safe. They have “gates” that prevent them from running if conditions are not permitted. For example, a function may be skipped if there is a pending restart, or the execution time is not in an allowed maintenance-window. This particular function will detect a logged in user and present a screen to them letting them know about the restart. They may be able to cancel the restart and it will run again later. If no logged in user, it just restarts the system. It also has a -DisableCancel and a -Force switch. -Force will bypass restrictions like the maintenance window.

17

I had a long conversation with the mcp server and my copilot agent. (Never thought I’d say that about AI Agents.)

I think I\we figured out how to get it to run scripts on assets. I’ll know more next week.

18

I am wondering if we need to be cautious about security with the new MCP server.

19

hey folks! we had a great session this week showcasing a service desk plugin one of our MSP partners built with us for Syncro/Claude connectivity.

some solid use cases and step-by-step walkthroughs, you can check out the recording and summary here: Syncro | Syncro MCP Workshop: Optimize Your Service Desk with Claude