Boat Monitoring with AI

Artificial Intelligence (AI) is revolutionizing the world. Having worked in technology for a long time, we enjoy using AI heavily ourselves. As passionate boaters and marine technology enthusiasts, we couldn’t help thinking about ways to apply it to boating!

We already monitor our two boats, leveraging AI heavily and in this article, we will be talking about how you can do it too. As a quick disclaimer, we will primarily talk about Saillogger, which is a solution we develop.

AI has many uses, but outside of work, the primary way we use it is through AI assistants: Instinct, Grok Bot, Muse, OpenAI’s dots, the newly announced Claude Duet, and others. By the time you read this, there will likely be even more. We use these assistants to help organize our daily lives, from booking concert tickets to helping with travel arrangements. My preferred assistant is currently Muse, while the Admiral sticks with her favorite, Grok Bot.

The most natural next step for us was to give those assistants the ability to monitor our boats. We now have a boat specialist agent that keeps an eye on our boats 24/7, even correlating boat data with events from our cameras, weather reports sent by Chris Parker (our weather router for the Caribbean), news about local events, hurricane reports for the Caribbean, and so on.

How It Works

Saillogger was born as a fully automated trip logging solution when we wanted to sail down the California coast from San Francisco to Mexico, but it has since grown to include many more capabilities, from remote monitoring to maintenance logging. It runs on a device aboard your boat, most commonly a Victron GX device or a Raspberry Pi, though other compatible computers connected to your boat’s network can work too.

You can explore its capabilities with a demo account on the website. What matters most here is its ability to monitor everything your boat reports via NMEA 2000 and automatically keep historical records for you.

More recently, we’ve added capabilities to Saillogger that let AI agents integrate easily with it, putting everything Saillogger can monitor and log for you at your AI assistants’ disposal with a single click. What’s more, we let AI assistants selectively configure monitoring settings in response to your prompts. They can help you figure out what additional things you can monitor on your boat and set up that monitoring for you. This is usually one of the trickier parts of setting up a boat monitoring and alarm system, but AI can be very helpful here.

The mechanism that connects AI assistants to Saillogger is called MCP (Model Context Protocol). It essentially allows an AI assistant to use the tools offered by a service provider. In this case, Saillogger offers a very rich set of tools that provide access to your entire trip history, statistics, and places you have anchored or docked, as well as your near real-time monitoring data and its history. All you need to do is tell your assistant to connect to Saillogger.

Connecting Your AI Assistant

Let’s see how it works in practice with Grok Bot. Just tell it to connect to Saillogger’s MCP server at https://saillogger.com/mcp.

This brings up an authorization window where you can approve the connection:

You can choose read-only access or allow the AI assistant to change your sensor settings and create alarms too. We will cover these in more detail further below and this also depends on how comfortable you are letting AI tweak your configuration, but in this example, we’ll allow it.

Exploring Your Boat’s Data

Once connected, Grok Bot will be able to retrieve near real-time data from your boat. Saillogger reports at one-minute intervals while your boat is online.

This gives a quick overview of the boat’s status, summarizing everything it can retrieve. The screenshot also happens to reveal that we’re taking a break from the boats and spending some time on land 😊

Grok Bot can now also access historical data and create visualizations or perform other analyses for you. Here is a simple example that gives you a sense of what is possible:

You can also let your AI assistant help you figure out which sensors you have and enable tracking for them in Saillogger.

Here is an actual example. We weren’t remotely monitoring solar output or the battery status of our RuuviTag Bluetooth sensors, which are placed in different parts of the boat. Grok Bot discovered these data sources and suggested monitoring them, as we’d already lost contact with one of the sensors in the port hull, likely because of a dead battery.

Just to add a bit of fun, let’s ask Grok Bot to summarize our last cruising season:

These are all excellent ways to use an AI assistant, especially when you use it more broadly and it can draw on your boating history, memories, journal entries, real-time data, and context while helping with your daily routines.

Having Your Assistant Keep Watch

But what matters even more to us boaters is having someone keep an eye on our boats while we enjoy a moment without thinking about them. So let's ask Grok Bot to keep an eye on our boat:

With a simple instruction like this, it will start checking on our boat regularly. What’s attractive about AI is that it already demonstrates the ability to analyze information in context and even apply judgment. This kind of monitoring can go beyond a basic rule-based alarm system that triggers when certain thresholds are crossed.

You can also provide more details about your boat, the equipment aboard, manuals, and so on, which can make its analyses more useful. To us, this feels close to having someone watch our boat 24/7, applying context and boat-specific know-how to what may be happening.

Here is a great example: Grok Bot noticed that the saloon sensor’s battery was running low and that something was drawing a fair bit of power at night. It did this without being told to do so or without explicit rules. It actually found that our sensor is about to die and noticed our Starlink, backup marina Wi-Fi connection, router, and Raspberry Pi, which together draw close to 100 W.

So far I’ve been demonstrating this with Grok Bot, starting with the initial setup, but you can also use other compatible AI assistants. As I mentioned earlier, the Admiral is an avid Grok Bot user, while I use Muse. My Muse is configured to watch both of our boats and keep me informed. If something happens, I hear from it. Here is a message I received today:

In this screenshot, you can see how it combines multiple sources of information: an email from Chris Parker (our weather router for the Caribbean), forecasts from Windy, and updates on our friends. Our friends on Mimi are on their way to the Bahamas, where we first met them. It was very nice to be automatically informed when they left the dock in the Chesapeake so we could connect with them!

When an Hourly Check Isn’t Enough

Taking a step back, what is already available here is amazing. However, one obvious gap in this approach is the frequency of the checks. In the example above, we asked Grok Bot to check things every hour. What if something goes wrong a few minutes after the last check?

First off, we don’t think it’s a good idea to rely solely on AI, or any form of remote monitoring for that matter, to detect a problem that could sink a boat at a moment’s notice. Those situations need dedicated protection aboard, such as an additional high capacity bilge pump triggering at high-water or an automatic fire suppression system, with notifications sent in parallel. That deserves some thought of its own, but regardless, a system that checks every hour or so will always have a gap between checks.

That’s why we also developed a mechanism that lets AI assistants define alarms for your boat and wakes them up when those alarms are triggered.

Here is roughly how it works: you or your AI assistant can define fixed rules that trigger when Saillogger detects a threshold being crossed. It could be wind rising above a certain speed, temperature dropping below a set value, or a bilge pump coming on. The AI assistant then gets notified alongside you, or instead of you if you prefer.

These rules normally notify you through email or a phone push notification in Saillogger. Now those alerts can also be delivered directly to your AI assistant through a webhook, allowing it to analyze the situation and take action within the permissions you’ve given it. Depending on the assistant’s capabilities, that might mean calling the marina or yard, or capturing video from compatible monitoring cameras. Your imagination and the capabilities of these AI agents set the limits, and those capabilities are growing very rapidly.

Alerts Set by Your AI Assistant

One key requirement for this approach is an AI assistant that can accept incoming webhooks. Not all assistants support this yet. At the time of writing, Muse doesn’t, though I’ve submitted several feature requests. Grok Bot does. Let’s see how it works in practice. Go ahead and tell Grok Bot to add a webhook for Saillogger:

It will provide a URL and a key in the panel on the right. Copy them into Saillogger’s Notifications settings, under Webhook Notifications:

Saving sends a test notification, which you can immediately see in Grok Bot:

Now you can choose which notifications to send directly to Grok Bot. You can turn individual notification categories on or off, or disable them all. For boat monitoring, keep “Monitoring Alerts” enabled so Grok Bot receives your alarm notifications.

Once webhooks are enabled, tell Grok Bot to set up alarms. You can be creative about how you want it to configure and respond to them. The good thing is that these settings don’t have to stay fixed. Grok Bot can adjust them based on where you are, your travel plans, the season, conditions, and so on. Here is an example:

To show this in action, we temporarily set the wind alarm threshold to 5 knots. And unsurprisingly, in less than 30 seconds, Grok Bot responded:

The layer of judgment that AI assistants can bring here is actually pretty impressive. It even guessed this was a test, without being told.

If something really is wrong, they can pull more real-time data, look at historical sensor readings, connect the dots, and put things in context. They can then take additional action, either automatically based on your prior instructions or by helping you diagnose and mitigate the issue. They can also visualize the data for you on demand. For example, my Muse is instructed to include trend charts whenever it flags a potential anomaly, so I can quickly assess whether something looks amiss.

You can also connect compatible assistants to messaging apps like WhatsApp, so you can literally text with your boat wherever you are. Here is how I communicate with Muse over WhatsApp:

We also have someone looking after our boat, and we communicate primarily over WhatsApp. We haven’t needed to do this yet, but we could ask our AI assistant to contact him directly when something needs checking. For example, it could ask him to check the dehumidifier or air conditioning if humidity is rising, or the shore power connection if the boat loses shore power.

What Comes Next

AI is opening up new possibilities in boating and marine technology. Every day, we hear about new boating software being developed, tested, and released. We expect this to accelerate and benefit all of us, as AI lowers the barriers for people to bring their ideas to life.

We’re excited about what this means for boating, both as users and as developers of new technologies. With this article, we wanted to show what you can already do at little or no additional cost if your boat is connected to Saillogger. Muse, for example, offers a free basic tier.

If you don’t use Saillogger yet but are interested, we have two pieces of good news. Getting started currently requires hardware aboard, such as a Victron GX device or a Raspberry Pi, and we know that can be a barrier. We’re developing our own hardware to make this significantly easier: a plug-and-play device that connects to NMEA 2000 and takes care of most things automatically after an initial setup. We’re currently testing prototypes and they work great. If you’re interested, let us know at [email protected] or [email protected].

You can also use a similar AI integration with our companion service, AIS Fleet, without having Saillogger aboard. AIS Fleet is free to use and has its own MCP server at http://aisfleet.com/mcp, available with its $2.99/month subscription tier. It doesn’t offer Saillogger’s onboard sensor integration or detailed monitoring capabilities, but it’s a great way to keep track of your own boat or your friends’ boats over AIS.

I don’t know about you, but I’ve always been fascinated by sci-fi stories where explorers venture into deep space with their intelligent ships and droids. We may not have the option to explore deep space yet, but I have to say I’m already pretty jazzed about being able to talk to my boat over text. I’m sure voice conversations will come pretty soon, for those who want them. On second thought, I might actually prefer to stick with text 😉