LLM Networking with MikroTik
blog.greg.technology101 points by gregsadetsky a day ago
101 points by gregsadetsky a day ago
I've been a long-time fan of Mikrotik. I even ran an ISP on it a fair while ago. I have a couple of archived GitHub projects if they are useful to anyone:
- Router OS Diff - Can diff two configs and give you the commands needed to bring the existing config up to date with the desired config. It's certainly not perfect, but a starting point of anyone needs something like this. [1]
- Netbox Routeros – A netbox plugin for updating the config of RouterOS devices directly from the Netbox interface. [2]
It has been many years since I touched these, but perhaps they will be of interest to someone.
Aside from that, I have had excellent experience with Mikroik. Everywhere from in-datacenter to it running in an off-grid hut on a mountainside. I've even heard reports of people finding rain streaming through a CRS and it just happily ticking along.
Diffs are a pain as the order of the export changes in a version by version basis
But using “export terse” is far easier to do a standard “diff”
MikroTik recently updated their documentation site from an Atlassian Confluence Wiki to much more AI-friendly Docusaurus here: https://manual.mikrotik.com/
Any page can be easily converted into Markdown by appending .md to the URL. I mention this because in my experience, the agent is much more accurate when it has access to the docs.
The page you linked to has a TOC for LLMs: https://manual.mikrotik.com/llms.txt
Neat, I'm kicking this off to the web team of the company I work for to see if they can quickly implement this for our documentation.
Finally the Confluence pages were really bad.
Thankfully at least LLMs could figure out things t help me setup SXT LTE 7
Thanks for sharing, this is a great improvement (also and especially for human use! Renders much better on mobile phones compared to Confluence).
It’s interesting to observe and build LLM-driven solutions in Networking.
The biggest challenges that most of us networking people have are around velocity (how fast we can build and scale networks) and how effectively we can operate them (avoid defects, fix them fast when something breaks).
LLMs are great in both areas. AI helps with deployment challenges by speeding up tooling development and the creation of workflows on orchestration platforms. A manual process step today, say - reserving an IP address in an IP DB — is automated the next day instead of on a backlog for years. This post is an example of that (config-gen/config-deploy).
Operations use-cases are more interesting, IMO, and address the “too many signals” problems that we face. Network substrate telemetry, overlay telemetry, service host metrics, service metrics, customer metrics, recent change data, prior alarms - the list goes on. Being a network operator is not for the faint of heart and is under-mentioned on high stress job lists. AI makes AMAZINGLY good network operations triage agents, since they are able to immediately process so many signals.
Exciting times!
>LLMs are great in both areas.
Nuance. LLMs are just going to report that they cant SSH to an endpoint, after delivering your vibeconfig, and throw it back to you to resolve connectivity. Your velocity with LLMs will stall at break fix every time.
>AI makes AMAZINGLY good network operations triage agents, since they are able to immediately process so many signals.
I have seen a lot of tokens spent on solutions that could have just been grafana.
I would expect LLMs to be especially excellent at configuring Mikrotik stuff, given MT publishes markdown reference docs for LLM ingestion, the full config without secrets can be dumped to one text file, and their cli commands are very stable between versions.
I switched recently to OpenWrt from MT, which code agents are also good at. I'd wager most issues are going to be related to the user not specifying what they want clearly enough. The translation from network concepts to RouterOS config is pretty 'fat-free', so there's not much room for hallucinations beyond syntax errors, which can be verified via the API.
LLMs as tech support is such a game changer. I have a fun complicated home network with a Mikrotik and a NAS and my desktop and laptop plus various radios and Raspis, and there's no way I could configure the extremely complicated RouterOS on my Mikrotik without AI. The speed is also helpful. Previously if my network went down I would have to wait until the next weekend to fix it because it would take me all day. Now I can get things back up in an hour or less.
Yes! Recently connected two disparate systems (ubiquiti and mimrotik) using their exposed API’s and a Claude session so that systems I have on either environment could talk to each other. I am not a network engineer so it was liberating to get my gear working together. That said it’s a work in progress and just today I noticed something weird that one of my computers can’t access Minecraft servers while the rest of my network can
Probably a routing issue. Shot in the dark would be that one of these routers is NATing traffic, and the other router doesn’t have a route to that NAT’d range.
Other shot in the dark, misconfigured bridging or similar where ARP isn’t getting forwarded and rewritten quite right
Or look for something mangling SRV records, Minecraft seems to use those more regularly than everything else.
What possibly could you even be talking about? Networking gear already works together due to standards.
LEGOs fit together because they have standards. That doesn’t mean they self-assemble.
Networking can be complex. Standards allow interoperability but they do not magically make everything work with no configuration.
You still would need to know how networking fundamentals and how network protocols work.
Just because standards like DNS, NAT64, OSPF, ARP, etc, exist doesn't means its easy to get these things to communicate.
Ubiquiti isn't exactly known for being the best in terms of standard adherence, especially with their historically week IPv6 support.
This is quite timely - just had some fun creating a OTLP/Grafana cloud exporter, which sends data directly without requiring intermediaries.
Super light, right now testing on my own set up (was using mktxp before):
https://github.com/jutaz/TikTelemetry
Comments/feedback welcome!
The important thing you need to do is specify which Mikrotik version you're using; apparently the syntax for some things changed between 6 and 7.
It does pretty well, but you need to iterate. I was trying to get it to disallow internet access for non-DHCP clients, and in the end there were so many limits to what was possible that it wasn't worth it. But it did it, and when I was testing I found them.
So like everything for best results you need to know what you're doing so you can test effectively...but it saves you from learning the syntax etc.
MikroTik is one of the only companies that sells a router without WiFi at a non enterprise price. Useful for me to completely turn the power off at night to a separate WiFi (router in Bridge mode).
> ... WinBox (which to my surprise is now cross-platform, and works quite well on Macs)
Winbox having a Linux and Mac version is really nice... I'll have to try the Linux one at some point... hopefully it works as good as the Windows version did under wine.
I've been doing some of this too. It's great just messaging claude in discord when I want to make a new device's IP static. I also put my mikrotik config sans passwords in (local only) version control, just in case.
And perhaps it's significantly easier on other routers, but I would not have gotten VLANs working without claude doing it for me...
I want something similar to this but for Ubiquiti. I don’t need anything fancy, just something that audits my home config and tell me if I’m doing something stupid, dangerous, or both.
Same. Recently Ubiquiti has been putting more and more into the local API, so this should be getting easier to do. The Home Assistant UniFi integration just recently has started moving from private undocumented API endpoints to the newer public API endpoints.
In other news, Meraki has an AI assistant feature now.
Wish I could configure pfSense via API or config files.
You can. Both pfSense [0] and OPNsense [1] have good API coverage.
[0] https://pfrest.org/api-docs/ [1] https://docs.opnsense.org/development/api.html
The main pfsense dev who exited the project is working on just that. Forget the name.
I only have the agent investigate directly. To actually configure the Mikrotik, I have the agent write a script that is aimed to be idempotent and then run the script. Investigation is fine, but the script acts as a memory of intent which I find useful. As agents get better, it can be a textual representation rather than a script, but for now that suffices.
> I have the agent write a script that is aimed to be idempotent and then run the script.
You can take this one step further and have the agent write Terraform configs [1]. I did this (including having the agent import all the initial resources from the live device), works great and is generally more robust than a script.
[1] https://github.com/terraform-routeros/terraform-provider-rou...
I originally wrote specific terraform providers (even one for just configuring an Ubuntu machine), but over time I found that TF is a bit too heavyweight for my use-cases. The shell script works well because state divergence can be investigated by the LLM. The slowness of state refreshing etc. does make a TF apply painful. For me at least.
How big is your TF state on your routers?
Yeah, TF is slow if you have thousands of configs and batching not integrated in the provider.
> As agents get better, it can be a textual representation rather than a script, but for now that suffices
I can’t see any reason to have agents do what a script can do. If the operation is deterministic then why pay every time it gets done? This is why MCP seems so pointless to me.
It's adaptive and can handle config drift if someone has altered the machine in the meantime between script invocations. Not required if you're disciplined, of course.
Mikrotik also has a safe mode for testing configuration
There is also a terraform provider. Not sure if there is a safe mode here. I normally test via ssh safe mode and import the changes afterwards.
Mactelnet is infinity useful, no need for terminal servers, you can bounce from your second mikrotik and reset the config no matter how screwed up your config.
Fantastic tip in this article to use version control (i.e. git) for config files, I hadn't thought of that.
I’ve been using ChatGPT to configure my mikrotik gear for about a year it’s pretty awesome. And the end result is well documented reusable scripts rather than my usual set of random stack overflow copy pastes and shitty inscrutable notes
A word of warning, my experience with mikrotik in the WiFi-6 space has been overwhelmingly negative, both in a single AP setup at home and on a corporate WiFi network managed by someone else.
Bugs, random drops, poor performance on some machines. It is not set and forget and I gave up on it.
Interesting that sending things like network configurations, keys, and credentials to external entities - which BTW are fueled by data - is considered "ok" now.
Is there anything here that suggests that? You are jumping to conclusions easily.
It's nowadays the default not to export any credentials when exporting MT configs, and even doing that yourself is trivial. Since the equipment is pretty packed with functionality and very flexible, benefits of LLM assistance is big.
As for "sending network configurations" it's a risk everyone needs to consider individually. Having that private may make attackers job harder, but security by obscurity ain't no replacement for a secure setup.
When I was setting up Wireguard server at home, the last step (after testing the setup) was the LLM giving me a set of commands to run manually to regenerate the credentials.
Once AI finds out you are located at localhost, it's all over for your boxen... RIP 192.168.xxx.xxx!
Yeah, the thought of having private keys and config slurped up by an LLM provider keeps me away from this kind of direct work.
Perfect job for a local LLM!
>One of the usual complaints about MikroTik has been its complex ui/configuration. In a sense, I don’t know if that’s true inasmuch as networking is complicated in itself
Really? Its standard point and click engineer stuff. The biggest issues with Mikrotik are the features not implemented in the gui, or the way config is interpreted between versions. Also the term of hardware support, and generally flaky code in general.
>The point I’m trying to make is yeah, networking can just be hard. I’ve been half-networking, amateur-ishly, for a while now - setting up networks for friends and friends’ offices, making cables, patching small panels etc. I almost certainly couldn’t pass an official “Certified Routing Engineer” cert - well, not without studying a lot (believe in yourself).
Ok so just a hobbyist perspective.
It seems like this article is just "Point an LLM at your mikrotik api, have fun"?
My issue with Mikrotik is that the UI puts all the complexity up front. A good UI should guide the user and reveal relevant information only as it's needed. Mikrotik doesn't do that.
For example, the most common reason I want to connect to my home router is to see what devices are connected, what their IP addresses are, and perhaps make their DHCP leases static. In a good UI that sort of common activity would be front and centre - in MikroTik it's buried under 3 levels of menu.
Under the IP menu is 26 alphabetically sorted options, of which I have to click "DHCP Server". Then the default page is to create a new DHCP server - why would I want to do that? How many users run multiple DHCP servers? I have to click on the "Leases" tab, and then I can see a list of my connected devices.
Every other home router I've used knows that users care about the connected devices, so show it front and centre.
> Really? Its standard point and click engineer stuff.
Really. UI is easy. CLI is easy. But system exposes everything to you. It doesn't hide any complexity. So you need to actually know what you are doing, as happily clicking randomly won't produce any reasonable result.
> Ok so just a hobbyist perspective.
No need to diminish those experiences. That's how most of us got into the job. And enterprise experience ain't exactly a guarantee of wide knowledge.