With the release of watsonx Code Assistant for Z v2.8.0, IBM is redefining what modern mainframe development looks like. This release introduces a brand-new agentic chat experience, enabling developers to collaborate with AI agents that understand intent, orchestrate multi-step tasks, and provide deep architectural reasoning across mainframe codebases. The new agentic capabilities empower WCA4Z to operate as an intelligent mainframe development partner—helping teams accelerate enhancements, modernization tasks, and architectural decisions at enterprise scale.

A New Agentic Development Experience

The new agentic chat interface brings goal-oriented reasoning to the forefront of the IBM Z development lifecycle. Instead of treating each action as an isolated request, the assistant can now:

• Interpret and refine developer goals

• Determine the appropriate products or capabilities to apply

• Break objectives into smaller steps when needed

• Ask clarifying questions to reduce ambiguity

• Combine understanding, explanation, planning, and implementation actions

This creates a more natural, guided development experience aligned with how developers work, especially in complex mainframe environments where tasks often touch multiple layers of the application architecture.

How Agentic Workflows Bring WCA4Z Capabilities Together

In earlier experiences, the products that make up the WCA4Z modernization lifecycle such as Code Generation, Code Explanation, and Understand operated independently, each delivering value within its own phase of development. With the introduction of the new agentic chat experience in v2.8.0, these capabilities can now be combined intelligently when a task spans multiple phases or requires deeper architectural context.

Agentic workflows often draw on these capabilities together to support development tasks that require both explanation and architectural awareness. For example:

• When deeper architectural understanding is required

• When planning should be generated before code is written

• When existing logic needs to be explained before updates or enhancements are made

• When code generation must account for relationships or dependencies not visible in the local workspace

In these scenarios, the agentic layer determines the sequence and combination of steps needed to fulfill the user’s request, allowing WCA4Z to deliver a more intelligent and contextually aware development experience, without forcing all capabilities into a single, universal workflow.

Specialized Agents

Z Code Agent — Explain, Generate, Modify, and Refactor

The Z Code agent acts as the hands-on development assistant for mainframe development, and related technologies. It helps developers:

• Explain existing program logic

• Generate new code aligned with enterprise standards

• Modify or extend existing applications

• Refactor large or complex modules in a controlled, consistent manner

When a requested update requires updates to multiple files in the workspace, the Z Code agent can invoke deeper architectural reasoning through Z Understand (via Z Architect mode) to produce an implementation plan and impact analysis, before generating code. This gives developers the architectural context needed to implement multi-file changes safely and accurately in large, complex applications.

Z Architect Agent — Deep Analysis with Z Understand

The Z Architect agent provides the architectural intelligence behind the agentic experience, powered by the Z Understand service. This capability is one of WCA4Z’s most important differentiators because it offers visibility far beyond what traditional code assistants can analyze locally.

Z Understand enables the agent to:

• Retrieve structural metadata from the entire application, not just the files in the workspace

• Perform dependency, and data-flow analysis

• Understand how a local change impacts remote components

• Map relationships across thousands of programs and copybooks

This is critical for enterprise-scale mainframe applications, which can be too large to fit into any single IDE workspace. WCA4Z overcomes that barrier by allowing the agent to reason globally, helping developers make safe, well-informed decisions regardless of application size or location.

Strengthening Code Quality Through Z Code Scan Integration

This release also deepens WCA4Z’s integration with the Z Open Editor MCP server, enabling seamless use of the Z Code Scan tool within agentic workflows. This capability expands the assistant’s role from helping developers understand and modify code to also ensuring that those changes align with enterprise coding standards.

Through Z Code Scan, WCA4Z can:

• Automatically scan application code for compliance with organizational rules and best practices

• Surface findings that may impact code quality, maintainability, or consistency

• Guide developers through resolving these findings

• Incorporate code scanning into multi-step agentic reasoning when appropriate

By bringing Z Code Scan into the agentic experience, this helps teams maintain high-quality codebases even as they accelerate development through AI-driven enhancements. This ensures that modifications—whether generated by the assistant or written by developers—remain consistent with established standards across large and complex mainframe applications.

Enterprise-Scale Differentiators: What Makes WCA4Z Unique

WCA4Z stands apart from other code assistants due to its ability to operate at true mainframe scale, supporting:

1. Extremely large programs (20,000+ lines)

WCA4Z uses intelligent selective analysis to navigate large program files without ingesting entire programs into LLM context, preventing token overload while retaining deep reasoning capability.

2. Workspaces containing thousands of files

To avoid overwhelming the model with large directory structures, v2.8.0 introduces Workspace Mapping, which uses Z Understand to generate documentation describing:

• Program relationships

• Key entry points

• Component interactions

• Directory and dependency structure

The AI uses this mapping—not raw file traversal—to locate relevant artifacts quickly and accurately.

3. Full-system architectural insight beyond the IDE

Because Z Understand has global visibility, the agent can reason about code not stored locally, a capability unmatched by standard code assistants that rely solely on workspace content.

Bringing It All Together: The Future of Agentic AI on Z

With the introduction of agentic chat in v2.8.0, watsonx Code Assistant for Z delivers a development experience that is:

• Goal-driven and intent-aware

• Architecturally intelligent through Z Understand

• Capable of orchestrating multi-step reasoning and planning

• Built for enterprise scalability across massive programs and repositories

This release marks a significant milestone in the evolution of AI-assisted mainframe development—establishing a new standard for how developers understand, modify, and modernize enterprise applications on IBM Z.