Soracom Agent: AI Assistant Aims to Guide IoT Projects From Planning to Operations
Dennis Knake
Posted On 3rd August 2026
Soracom has introduced Soracom Agent, an AI agent designed to support companies from initial requirements analysis through the ongoing operation of an IoT project. The service is launching as a technology preview. WeSpeakIoT asked Soracom which language model powers it and how the company handles data protection for European customers.
With Soracom Agent, the Tokyo-listed company (TSE: 147A) extends its existing AI services: Soracom Query, which enables natural-language data analysis, and Soracom Flux, a low-code builder for IoT applications and automated workflows.
Users are meant to describe their project goals to the agent via chat or voice. Capabilities listed in the product announcement include operating Soracom Cloud Camera Services (SoraCam), analyzing device data and images, remotely controlling devices through natural language, and automatically building applications with Soracom Flux. Technically, the agent accesses these through the platform’s API, command-line interface (CLI), and the Model Context Protocol (MCP), an open standard that lets AI systems operate external tools and services.
Each agent reportedly runs in its own isolated container and features what Soracom calls Project Memory, which retains business context and system configuration over time. According to the announcement, customer-provided data is neither shared with other users nor used to train general-purpose AI models.
Which language model powers the agent?
In written responses to WeSpeakIoT, company to co-founder and CTO Kenta Yasukawa said it uses multiple language models across its AI services. Yasukawa pointed to a partnership with OpenAI announced in 2025 but did not name the specific model behind Soracom Agent. Customers currently cannot select which model is applied. Soracom may consider offering multiple model options during the technology preview if demand emerges.
What about data centers and GDPR compliance?
On data processing, Yasukawa said Soracom uses multiple AWS regions for inference and typically routes European customers through the AWS Frankfurt region. The Soracom platform is generally GDPR-compliant, he said; the specific routing for Soracom Agent itself is still being actively evaluated during the ongoing technology preview.
Asked about the technical nature of the separation, Yasukawa specified that the agent orchestrator runs in one or more containers dedicated to each customer, keeping context and project knowledge separated. He left open whether this means separate compute infrastructure or separation on shared infrastructure.
Is there a test environment available?
According to Soracom, there is currently no publicly accessible sandbox environment. Instead, the service is enabled individually based on each customer’s use case. A broader test environment may become available once the product moves from technology preview into a beta phase.
For now, Soracom Agent remains an offering enabled on a per-account basis after individual review. Whether and in what form the service becomes more broadly available will likely become clear once it moves into its beta phase.
