About the Agent Centric Development Cycle

The Agent Centric Development Cycle helps you ensure code generated by agents is verified and issues are fixed before the code reaches your repository.

Overview

The Agent Centric Development Cycle (AC/DC) is Sonar’s framework for building software in an AI-driven world, where agents generate most of the code and humans govern quality. It defines a continuous 4-stage loop for every AI-generated change:

Guide → Generate → Verify → Solve

• Guide provides project context to agents before code generation.

• Generate is where the AI agent writes code.

• Verify analyzes the generated code.

• Solve fixes issues found during verification.

This workflow helps agents produce code that aligns more closely with project standards, catches issues earlier, and shortens the review-to-fix cycle.

The diagram below shows where AC/DC features fit in this cycle.

Guide

Context Augmentation provides project-specific context—coding guidelines, architectural information, and codebase context—to the AI agent before code generation. This reduces avoidable violations and inconsistent code patterns.

Generate

Your AI agent generates or updates code using the context and constraints gathered in the Guide phase.

Verify

Agentic Analysis analyzes generated code during the agent workflow. The agent can use the analysis results to refine the code before review. This surfaces quality and security issues before they reach pull request review or CI.

AI Code Review posts inline review comments, change summaries, and on-demand walkthroughs and architecture diagrams directly on pull requests.

Solve

The Remediation Agent generates fix suggestions for issues found in pull requests and in the backlog. AI CodeFix generates fix suggestions for individual issues detected by SonarQube. This reduces manual rework and supports incremental cleanup of existing issues.

Get started

See Features for product details and How-to guides for setup and usage.