Your Fleet Data Inside Your Team’s AI Tools: the Soracom Query Connector for Claude and ChatGPT

TL:DR

• Soracom Query’s new MCP connector is live in the Claude and ChatGPT directories, letting your AI assistant query connectivity, billing, and device data directly from your Soracom account.
• The bigger win isn’t the reports themselves, it’s that your fleet data now shares context with your other tools, so an assistant can cross-reference usage anomalies against deployment records or logs in one conversation.
• It’s read-only for analysis, so ask clear, well-scoped questions, verify any numbers headed for a customer or renewal decision, and watch your query volume against your billing plan.

---

I have spoken before about how Soracom Query can be used to answer the most common questions within a business about your company’s IoT fleet. 

Questions like: 

• Which SIMs went up in cost last month and why?
• Which devices have not been online since the summer?
• How much of this fleet is still sitting on a network that is about to be switched off?

These are all good questions, and until Soracom Query came along, answering them properly meant someone like me writing SQL against the data, or exporting a CSV and living in a spreadsheet for an afternoon.

Soracom Query has been the answer for a while, but a new update released this week has made that solution more accessible. The Soracom Query MCP connector is now live in the Claude directory and ChatGPT, which means your own AI assistant can query your fleet data directly in the same conversation you are already working in.

---

What Has Launched

MCP, the Model Context Protocol, is the emerging standard for letting an AI assistant use external tools and data sources rather than relying on training alone. A connector is just a server that speaks to it.

Earlier this year, Soracom released Soracom Query MCP, hosted here . By adding this connector to your Claude or ChatGPT and logging in to your Soracom account, your AI tools now have access to the wealth and breadth of knowledge of your Soracom instance.There is nothing to install,  no key to paste into a config file. You simply add the connector, approve access, and from then on the assistant can query the data in your account. Authorization is tied to your Soracom login, and it is based on an Auth token that you will need to refresh. 

---

Why This is More Than an AI Gimmick

By empowering your tools to access Soracom Query, your assistant can now answer questions about:

• Connectivity: session history, which country and network each SIM is on, cell tower detail, connection and disconnection events, data usage.
• Billing: monthly billing history including amounts and plan detail.
• User data: the device data you have accumulated in Harvest Data as time series, plus metadata on files in Harvest Files.

That is the platform’s own view of your fleet, not a sample, and it is scoped to your account. So the questions become conversational rather than analytical projects:

• Which SIMs used more than three times their usual data last month, and what networks were they on when it happened?
• Show me every device that has not created a session in thirty days, grouped by country, before we decide what to do with them at renewal.
• Which of our devices are still attaching to 2G only, and where are they?
• Plot reconnection frequency by site for the last week and tell me which ones look abnormal.

I do versions of all of these by hand, regularly. The 2G one in particular is what a lot of European fleets should be running right now, with France ending 2G across all three operators in 2026. Being able to ask it directly, and then immediately ask the follow-up, is a different working pattern from filing a request and waiting.

---

The Part I Think Teams Will Get the Most From

The real advantage of Soracom Query is not the direct reports, you could likely pull those before. No, the real advantage is that your fleet data is now sitting in the same context as everything else your assistant can see, meaning all the manual steps and time it took to make the most of your data are a thing of the past.

An agent that can read your Soracom data and your own systems in the same conversation can join them up. Cross-reference the SIMs with abnormal usage against your device deployment records and tell you which customer sites they belong to. Take the session history for a device and read it next to the device’s own application logs. Pull the numbers and then write the summary for the Monday operations meeting, with the chart, without anyone copying anything between tools.

It also pairs well with the Soracom Knowledge MCP server, which is a free, read-only endpoint over our documentation and API reference. Knowledge tells the assistant how the platform behaves, Query tells it what your fleet is actually doing. Together you get an assistant that can spot an anomaly and then explain the mechanism behind it rather than guessing at one.

---

A Few Practical Notes

Give the assistant clear questions. A vague prompt over a large dataset produces a vague answer, in exactly the same way a vague prompt produces mediocre SQL. For truly useful insights, define the time window, define  the grouping, define  what “abnormal” means to you.

Check the results that matter. Query’s SQL assistant is good, and the connector is the same engine underneath, but a number that is going into a customer conversation or a renewal decision deserves a second look at the query that produced it. That is not a criticism of the tool, it is how anyone should treat generated analysis, all LLM’s can sometimes cause hallucinations, and these will need to be checked if doing reporting based on its output. 

Watch your query volume. Query plans differ in how SQL execution is counted and charged, and an agent exploring a question will happily run several queries where a person would run one. It’s well worth understanding how your plan counts before you point an enthusiastic agent at it.

Treat it as the analytics surface. This connector is for reading and analysing what your fleet has done. It is not able to change SIM states or push configuration.

---

Where This is Heading

The shift here is small to describe, and quite large in practice. Fleet data stops being something you request from whoever knows the query language, and becomes something anyone on the team can pull from the tool they already have open. The SIM data, the session history, the billing, the sensor readings, all reachable from the same conversation where the rest of the work is happening.

If you are already on Soracom Query and using AI agents, add the connector and try the most recurring question you have. 

---

If you are not using Query, please reach out to our team and we can discuss a trial, and what it can do for you and your team. This connector is a door into the Query service rather than a service in its own right.