Profound Logic Adds MCP To IBM i AI Tool
July 28, 2025 Alex Woodie
The world of AI is moving extremely fast right now, with new models, techniques, and technologies emerging every week. One of the newest bits of AI tech to catch on is Anthropic’s Model Context Protocol (MCP), which has quickly become a defacto standard for connecting data sources to AI models. Profound Logic, which has been on the cutting edge of AI, is the first IBM i vendor to bring MCP support to the server and the Db2 for i database.
Anthropic quietly launched Model Context Protocol in late November 2024 as a way to enable large language models (LLMs), such as its own Claude offering, to exchange data with databases, file systems code repositories, integrated development environments (IDEs), and other tools. By making MCP open, Anthropic hoped that it would eliminate the need for AI engineers to build custom connectors and APIs for each data source and each tool, which the company described as the “N×M” data integration problem (where “N” represents LLMs and “M” represents tools).
In early 2025, MCP started catching on. Other LLM providers, such as OpenAI and Google, pledged to support the protocol, which includes primitives that define what MCP clients and MCP servers can offer each other across categories like tools, resources, and prompts. Today, MCP is viewed in the AI industry as a key standard, akin to the ubiquitous USB-C cords and ports that billions of humans use to connect and charge their mobile devices.
MCP has caught the eye of IBM, which adopted the fledgling standard in its IBM watsonx suite earlier this year and now includes an MCP server with watsonx.ai. MCP is also turning some heads at the IBM lab in Rochester, Minnesota. While MCP isn’t part of the watsonx Code Assistant product that IBM i Chief Architect Steve Will is spearheading, “At some point we will add an MCP server for some things” Will told IT Jungle at the POWERUp 2025 conference in May.
With its June 2025 announcement, Profound Logic became arguably the first IBM i software vendor to add support for MCP. Profound added MCP support to Profound AI, which is a new framework it launched in late 2023 to streamline development of generative AI applications that work with IBM i applications and data.
In the beta versions of Profound AI, the vendor took responsibility for figuring out which APIs and protocols to use to connect IBM i apps and data to the various LLMs. Now that it has adopted MCP, all that technical complexity is handled automatically in the protocol. That will be a boon to Profound AI users, says Profound Logic CEO and Founder Alex Roytman.
“You don’t have to worry about the plumbing, basically. That’s a lot of it,” Roytman says. “Now you just kind of get down to what is the business problem that you’re trying to solve.”
Connecting IBM i resources to an LLM in Profound AI is as simple as choosing which database tables you want to use, or picking which RPG or COBOL programs to call, he says. It’s all done in a point-and-click manner in the low-code Profound AI IDE.
ProfoundAI connects a range of LLMs to IBM i data and applications.
Profound claims that supporting MCP will reduce the time it takes to build generative AI and agentic AI workloads from months to just days. With MCP functioning as the USB-C port, customers can pick and choose which LLM to use – Claude, GPT, Gemini, Llama – without incurring additional work.
Roytman says the majority of early Profound AI implementations have revolved around business intelligence use cases, where users want to ask questions of their data. In this mode, Profound AI functions a bit like an AI-powered data analyst by connecting the LLM and Db2 for i database. The user provides an English-language question, such as “What were the top sales people for the Southwest region in June”? The LLM will translate that into a SQL statement, Profound AI will route that statement to Db2 for i for execution, and then route the response back to the LLM, which translates the results into plain English for the customer to consume.
Profound AI customers are adding little pops to their screens that allow users to bring an LLM into play with whatever data they have open, Roytman says.
“It’s basically like an assistant that you can ask a question and it will do something on your behalf against the data that you’re working with,” he says. “You can even say, in my current view, can I get that as an Excel spreadsheet? And it will do it for you on the fly.”
Question and answer systems are an easy way for IBM i shops to dip their toes into the AI waters, but there are deeper waters to explore, such as using Profound AI to call IBM i programs. For instance, a user could use the Profound AI low-code environment to define the steps that he wants to occur, such as fetching a piece of data, generating an email, and then sending the email, Roytman says. One customer used Profound AI to automate the password reset process, he says. You could even use it to monitor an inbox, and respond
“Those are all possibilities,” he says. “That’s where a tool like Profound AI with the MCP support can come in . . . and connect to your actual source of truth, your IBM i data.”
Most ProfoundAI customers start with a question-and-answer application that lets them query Db2 for i data using plain English.
Now that MCP has eliminated a host of complex technology issues that were slowing AI adoption among IBM i customers, Roytman is hopeful that IBM i shops will move forward more quickly with AI. He is advocating for IBM i prospects to trust the vendor to help them brainstorm possible uses of the technology, and to engage with Profound for technical services.
“We want to basically be a solution instead of just a product,” Roytman says. “We want to be a partner, so to speak, not just a vendor.”
The IBM i space isn’t known for being early adopters of new technology, and that’s once again the case with AI. While Roytman admits to expecting adoption to be a little faster than it has been, he’s confident that with the right product and the right partner, IBM i shops will be able to move forward with their AI initiatives.
To get started with AI, Roytman encourages IBM i shops to pick a relatively easy project, such as enabling question and answering on Db2 for i data. Once customers have a little bit of experience, they’ll better understand the potential of AI technology in their shop and be better positioned to move forward, he says.
“We try to bring you into the future,” Roytman says. “Part of what we try to do is introduce these concepts. It does seem like the typical IBM shop is a little behind, but that’s going to be the future. And if they want to be competitive, if they want their organization to be competitive, you’ve got to start looking at that stuff and making a plan for implementing it.”
