AI for Mainframe Development: From AI Coding Assistant to Governed Deployment on z/OS

Yes. BMC AMI DevX Code Pipeline lets supported AI coding assistants invoke mainframe pipeline operations from inside the developer's IDE session. The assistant proposes the change, and on the developer's approval, Code Pipeline runs Generate, Promote, Deploy, or Regress on z/OS. COBOL, PL/I, JCL, and Assembler are all in scope. The assistant does not touch the mainframe itself. It calls the pipeline, and the pipeline executes under the lifecycle controls and audit trail already in place.

BMC AMI DevX Code Pipeline supports assistants that implement the Model Context Protocol, including GitHub Copilot, Anthropic's Claude Code, Cursor, and OpenAI Codex CLI. The integration is protocol-driven rather than vendor-specific, so an organization that standardizes on a different MCP-compatible assistant connects to Code Pipeline the same way. Developers reach Code Pipeline from VS Code and Eclipse through BMC AMI DevX Workbench as well.

Through the pipeline that already governs human-driven changes. In BMC AMI DevX Code Pipeline, an action invoked by an AI coding assistant runs the same path as an action invoked by a developer: the same lifecycle stages, the same promotion gates, the same audit trail. Compliance teams read one record covering both. Organizations in financial services, insurance, and healthcare need that equivalence before an assistant reaches the production path at all.

No. BMC AMI DevX Code Pipeline operates as the source code manager with feature branching, or as the build and deploy layer under Enterprise Git in GitHub, GitLab, Bitbucket, or Azure Repos. Assistant-invoked operations run the same way in either configuration, because the assistant calls the Code Pipeline API rather than the repository. Teams that have not adopted Git workflows get the same AI-assisted delivery path as teams that have.

The Model Context Protocol is an open standard that lets AI assistants call external systems through a consistent interface. For mainframe CI/CD, it means an assistant can drive pipeline operations without custom integration code written per assistant. BMC AMI DevX Code Pipeline implements it through the Code Pipeline MCP Server, which translates an assistant's request into a Code Pipeline operation over the existing REST API. The operation executes on z/OS under the same governance as any other pipeline action.

No, and that is the design. BMC AMI DevX Code Pipeline keeps a human in the loop. The assistant proposes the operation and the developer approves it before Code Pipeline executes. Promotion to higher lifecycle stages carries its own approval gates, with approvals issued from web or mobile. Autonomous deployment to production is not an appropriate posture for regulated mainframe workloads, and Code Pipeline is not built to offer it.

Reading source lets an assistant explain a program, suggest a change, or answer a question about the codebase. Running the pipeline lets it compile that change, promote it through lifecycle stages, deploy it, and leave an audit record. The second requires the assistant to reach build and deployment operations, not just the repository. BMC AMI DevX Code Pipeline exposes the full set: Generate, Promote, Deploy, and Regress, each under the lifecycle governance the pipeline already enforces.

BMC AMI DevX Code Pipeline integrates with existing mainframe DevOps tools and workflows through REST APIs, DevOps Extensions, webhooks, and command-line interfaces. Jenkins, GitHub Actions, Azure DevOps, and other orchestrators can drive pipeline operations on z/OS, while supported AI coding assistants connect through the Code Pipeline MCP Server.