Dortmund – November 09, 2026 -- Task Force IT-Consulting GmbH has released CMOne Agentic, an extension of its CMOne change-management system that lets AI agents develop, compile and unit-test changes to legacy IBM i applications. The product has been available since the fourth quarter of 2026 and is already in production use at fashion company LERROS.
CMOne MCP server gives AI agents application context without system access
The technical core is the CMOne MCP server, built on the Model Context Protocol, an open standard for giving AI systems access to information and tools. Through it, agents see the links between programs, their dependencies, the commands needed to edit them, and ongoing tasks, projects and changes already in quality assurance. The agents never access the IBM i system directly. All sources, information and commands pass through the MCP server.
Agents plan, code and test; developers keep release authority
In agentic coding, a developer sets a task and the agent plans and executes the steps. It analyzes existing programs, makes changes, compiles them and generates and runs unit tests with the iRPGUnit tool. If tests fail, it corrects its changes and repeats the check.
Task Force names three use cases: modernizing older RPG programs, converting existing data structures into SQL tables, and splitting large programs into smaller components. Agents can also document business processes and data models. Several specialized agents can work in parallel, with one analyzing program structure, a second making changes and a third reviewing the result.
AI-generated changes pass through the same review and approval process as those from human developers. Agents handle check-out, development, compilation and unit tests. Developers remain responsible for integration tests, promotion to the next stage and release into production. Development, test and production environments stay separate, and every change is documented and protected by defined permissions.
Tool-agnostic design supports Claude, GitHub Copilot, ChatGPT and IBM Bob
CMOne Agentic works with Claude, GitHub Copilot, ChatGPT and IBM Bob, so companies do not have to redesign development processes for each AI tool. A CMOne extension for Visual Studio Code brings core CMOne functions into that environment. For Profound UI users, Profound CoderFlow can build screen interfaces agentically, with the MCP server supplying context.
Retiring developers drive demand for AI-assisted IBM i maintenance
Task Force says experienced IBM i developers are retiring and qualified successors are hard to find. The applications they maintain often run accounting, warehouse management and logistics. CEO and founder Andreas Strietholt warns that AI can produce in minutes code that once took days, but that without an understanding of the existing application it can introduce errors costing many times as long to fix. The company has served mid-sized and larger customers in German-speaking countries since 1999.
CMOne Agentic requires CMOne version 6.1.29 or later. It is aimed at existing CMOne customers and at companies adopting CMOne for the first time. Customers decide which development tasks to hand to AI.